Graduation Committee:
Prof. Dr. A.V. Metrikine
Dr. Ir. M.A.N. Hendriks
Dr. Ir. E.M. Lourens
Ir. W.P. Kikstra
Author
:
Student ID :
Date
June 2013
by
Luis Fernando Sirumbal Zapata
June 2013
Graduation Committee:
Prof. Dr. A.V. Metrikine (chairman)
Section of Structural Mechanics, Faculty of Civil Engineering and Geosciences, Delft University of Technology
PREFACE
This research work constitutes the graduation project of my studies in the Master of Science in Civil
Engineering program at Delft University of Technology, Faculty of Civil Engineering and Geosciences.
During the last two years in TU Delft, I followed the structural mechanics specialization of the
structural engineering track.
This research project about the Hybrid FrequencyTime Domain (HFTD) method for nonlinear
structural dynamics was sponsored and developed in cooperation with TNO DIANA Finite Element
Software Company for Civil Engineering.
I am deeply grateful to Wijtze Pieter Kikstra, Senior Software Development Engineer at TNO DIANA,
because of his unconditional commitment with this project. He is my mentor and friend, and without
his participation the results obtained in the project would not have been as good as they are.
I would like to thank Dr. GerdJan Schreppers, CEO at TNO DIANA, for giving me the opportunity of
work in this project. Special thanks to: Jonna Manie, Software Development Engineer, which
collaboration was crucial in the most important stages of the project; and Maziar Partovi, Advanced
Application Engineer, who was always friendly and eager to talk about life. In general, many thanks to
all the TNO DIANA people with whom I worked and spent enjoyable time during the last eight
months.
From TU Delft I would like to thank my supervisor, Prof. Dr. Andrei Metrikine of the Section of
Structural Mechanics, for his valuable guidance during the graduation committee and personal
meetings we had during the last months. I would also like to acknowledge the important participation
of the graduation committee members, Dr. Max Hendriks and Dr. ElizMarie Lourens. And finally I
want to express my gratitude to Prof. Lambert Houben who guided me through all the graduation
process.
I would like to dedicate this work to my home country university, National University of Engineering
(UNI), in Lima, Peru. I am particularly grateful to the people of some institutions, like the Civil
Engineering Faculty Research Institute (IIFICUNI) and the National University of Engineering
Foundation (ProUNI), who believed in me and supported my studies at TU Delft.
Finally, I proudly dedicate my work to my family, parents and brother, and beloved persons in Peru.
Without their love and constant support it would not be possible for me to be here writing these lines.
L. Fernando Sirumbal Z.
Delft, June 2013
vi
ABSTRACT
This research work studies the fundamentals of the Hybrid FrequencyTime Domain (HFTD) method
for dynamic analysis of structures, applied to the seismic analysis of damreservoir interaction
systems.
The theoretical consistency and accuracy of the HFTD method, based on the frequency domain
Fourier analysis, is demonstrated through all the parts of the thesis, but particularly in Chapter 5 for
linear and highly nonlinear single degree of freedom (SDOF) models subject to several types of
loading, and in Chapter 7 for a 2D damreservoir interaction finite element model subject to an
earthquake ground acceleration loading.
On the other hand, a Finite Element Method (FEM) formulation of damreservoir interaction problems
is presented, being demonstrated the frequencydependence nature of this type of systems. The fluid
compressibility and boundary conditions of the reservoir are analytically identified as the frequency
dependent properties of the system.
In this context, the HFTD method is proposed as an ideal method for the dynamic analysis of
frequency dependent systems with nonlinear behavior, being the damreservoir interaction problem
taken as a case study to prove this hypothesis.
Therefore, one important part of the research work is the programming and implementation of the
HFTD method in MATLAB for SDOF models, and in DIANA Finite Element Software for multi degree of
freedom (MDOF) general models. In both cases, the results obtained with the HFTD method are
compared with the solutions of analytical or numerical methods in the time domain, like the wellestablished Newmark method.
Through the entire thesis, special attention is given to the study of the following topics:
For MDOF systems the HFTD method is compatible with the Mode Superposition method, which
reduces considerable the size of the problem and therefore the computational time of analysis.
However, the Mode Superposition method also showed to have a great influence in the accuracy of
the results, especially when the HFTD method is applied to nonlinear MDOF systems.
vii
Additionally, the influence in the HFTD performance of other important implementation issues, like the
time segmentation approach, the time step size and the Fourier Transform period, is explained and
numerically demonstrated.
One important conclusion of the research work is that the best performance of the HFTD method, in
combination with the Mode Superposition method, is obtained when it is applied to MDOF systems
with frequency dependent properties and linear or mild nonlinear behavior.
From the research work it can also be concluded that these frequency dependent properties actually
have an impact in the response of the damreservoir interaction system. However, it was not possible
to outline any definitive conclusion about the relative importance of their inclusion in the systems
response.
Further case studies, especially 3D damreservoir interaction models, should be analyzed using the
HFTD implementation in DIANA developed in this research work, with the objective of describing more
clearly the relative importance of the frequency dependent properties in the linear and nonlinear
response of this type of structures.
Finally, it is concluded that the HFTD method shows promising perspectives for its application in the
academic and professional practice, especially for the seismic analysis of fluidstructure and soilstructure interaction systems, or any other area for which the research of dynamic nonlinear behavior
of frequency dependent systems becomes every time more important.
viii
Table of Contents
TABLE OF CONTENTS
PREFACE....v
ABSTRACT......vii
TABLE OF CONTENTS......ix
LIST OF FIGURES......xiii
LIST OF TABLES..xix
LIST OF SYMBOLS..xxi
1.
2.
3.
INTRODUCTION.1
1.1
1.2
1.3
2.2
Direct method...7
2.3
2.4
Fourier analysis...9
2.4.1.
2.4.2.
2.4.3.
3.1.2.
3.1.3.
3.1.4.
3.1.5.
3.1.6.
3.1.7.
3.1.8.
3.1.9.
3.2
4.
3.2.2.
3.2.3.
3.2.4.
4.2
4.3
Implementation issues..32
4.4
5.
Table of Contents
4.3.1.
4.3.2.
4.3.3.
4.3.4.
4.3.5.
4.3.6.
Convergence criteria41
4.4.2.
4.4.3.
4.4.4.
5.2
5.3
5.2.1.
5.2.2.
5.4
6.
5.3.2.
5.3.3.
6.2
6.3
6.4
7.
Table of Contents
6.3.1.
6.3.2.
6.3.3.
6.4.2.
6.4.3.
7.1.2.
Fluidstructure interface..115
7.1.3.
7.1.4.
Foundation modeling117
7.1.5.
7.2
7.3
7.3.2.
Eigenanalysis..121
7.3.3.
7.3.4.
xi
8.
Table of Contents
Conclusions...............................................................................................................................145
8.2
Recommendations....147
REFERENCES..149
APPENDIX A: FLOWCHART OF THE HFTD METHOD IMPLEMENTATION IN DIANA.151
A.1.
A.2.
A.3.
A.4.
B.2.
B.3.
B.1.2.
B.2.2.
B.3.2.
xii
List of Figures
LIST OF FIGURES
Figure 2.1: Rotating force, displacement, velocity and acceleration vectors in the complex plane.........................6
Figure 3.1: Fluid and solid domains, fluidstructure interface and boundary conditions......................................13
Figure 4.1: Flowchart of the HFTD method general solution procedure.............................................................33
Figure 4.2: El Centro NS earthquake ground acceleration signal.....................................................................35
Figure 4.3: Spectrum of the ground acceleration amplitude for Fourier
period equal to the loading duration (
)...................................................................................36
)..............................................................................36
List of Figures
xiv
List of Figures
List of Figures
List of Figures
Figure 7.28: Incompressible fluid linear elastic response. Dams crest velocity Xdirection................................133
Figure 7.29: Incompressible fluid linear elastic response. Dams crest acceleration Xdirection.........................134
Figure 7.30: Incompressible fluid linear elastic response. Hydrodynamic pressures at different time steps........135
Figure 7.31: Incompressible fluid concrete cracking nonlinear
response. Dams crest displacement Xdirection............................................................................................136
Figure 7.32: Incompressible fluid concrete cracking nonlinear response. Dams crest velocity Xdirection.........136
Figure 7.33: Incompressible fluid concrete cracking nonlinear
response. Dams crest acceleration Xdirection.............................................................................................137
Figure 7.34: Incompressible fluid reservoir nonlinear response.
Dams concrete cracking pattern at different time steps................................................................................138
Figure 7.35: Incompressible fluid reservoir nonlinear response.
Displacement and principal strain fields at time step 2.40 s...........................................................................139
Figure 7.36: Compressible fluid with radiation boundary. HFTD nonlinear
solution for the dams crest displacement in Xdirection................................................................................140
Figure 7.37: Compressible fluid with radiation boundary. HFTD nonlinear
solution for the dams crest velocity in Xdirection.........................................................................................140
Figure 7.38: Compressible fluid with radiation boundary. HFTD nonlinear
solution for the dams crest acceleration in Xdirection..................................................................................141
Figure 7.39: Influence of the damreservoir interaction and fluid compressibility in
the concrete cracking nonlinear response. Dams crest displacement Xdirection............................................141
Figure 7.40: Influence of the damreservoir interaction and fluid compressibility in
the concrete cracking nonlinear response. Dams crest velocity Xdirection....................................................142
Figure 7.41: Influence of the damreservoir interaction and fluid compressibility in
the concrete cracking nonlinear response. Dams crest acceleration Xdirection.............................................142
Figure 7.42: Influence of the damreservoir interaction and fluid compressibility in the
nonlinear response. Dams concrete cracking pattern at different time steps.................................................143
xvii
List of Figures
xviii
List of Tables
LIST OF TABLES
Table 5.1: Summary of the HFTD method for SDOF nonlinear systems...........................................................51
Table 5.2: Iterative procedure of the HFTD method for SDOF nonlinear systems..............................................52
Table 5.3. Coefficients for the recurrence formulas of the Direct Integration Method.........................................58
Table 5.4: Summary of the Newmark method for SDOF nonlinear systems.......................................................61
Table 5.5: Summary of the Modified NewtonRaphson iterative procedure.......................................................62
Table 5.6: Numerical data of Example 4.2 (Craig and Kurdila, 2006)................................................................63
Table 5.7: Parameters of the SDOF linear system subject to harmonic loading..................................................63
Table 5.8: Analytic solution of SDOF linear system subject to harmonic loading................................................63
Table 5.9: HFTD parameters for the SDOF linear system subject to harmonic loading.......................................64
Table 5.10: Numerical data of Example 4.2 (Paz and Leigh, 2004)...................................................................66
Table 5.11: Parameters of the SDOF linear system subject to blast loading......................................................66
Table 5.12: Direct Integration method coefficients for
the SDOF linear system subject to blast loading.............................................................................................66
Table 5.13: HFTD parameters for the SDOF linear system subject to blast loading............................................66
Table 5.14: Numerical data of Example 4.7 (Paz and Leigh, 2004)...................................................................69
Table 5.15: Parameters of the SDOF linear system subject to impulsive loading................................................69
Table 5.16: Direct Integration method coefficients for the SDOF linear system subject to impulsive loading........69
Table 5.17: HFTD parameters for the SDOF linear system subject to impulsive loading.....................................69
Table 5.18: Numerical data of Example 4.9 (Paz and Leigh, 2004)...................................................................72
Table 5.19: Parameters of the SDOF linear system subject to earthquake ground acceleration...........................72
Table 5.20: Direct Integration method coefficients for the SDOF
linear system subject to earthquake ground acceleration................................................................................72
Table 5.21: HFTD parameters for the SDOF linear system subject to earthquake ground acceleration................72
Table 5.22: Numerical data of the SDOF system for resonance case studies.....................................................74
Table 5.23: Numerical data of case study A: Excitation frequencies close to resonance.....................................75
Table 5.24: HFTD parameters for case study A...............................................................................................75
Table 5.25: Numerical data of case study B: Resonance excitation frequency with very low damping ratios........78
Table 5.26: HFTD parameters for case study B...............................................................................................78
Table 5.27: Numerical data of Example 5.5 (Chopra, 2007).............................................................................80
Table 5.28: Parameters of the SDOF system with stiffness nonlinearity subject to sine pulse loading................80
xix
List of Tables
Table 5.29: Newmark method (average acceleration) parameters for the SDOF
system with stiffness nonlinearity subject to sine pulse loading......................................................................81
Table 5.30: HFTD parameters for the SDOF system with stiffness nonlinearity subject to sine pulse loading......81
Table 5.31: Numerical data of Example 8.15 (Humar, 2002)............................................................................84
Table 5.32: Parameters of the SDOF system with stiffness nonlinearity subject to blast loading........................85
Table 5.33: Newmark method (average acceleration) parameters for the
SDOF system with stiffness nonlinearity subject to blast loading.....................................................................85
Table 5.34: HFTD parameters for the SDOF system with stiffness nonlinearity subject to blast loading...............85
Table 5.35: Parameters for the definition of the nonlinear damping force.........................................................88
Table 5.36: Newmark method (average acceleration) parameters for the SDOF
system with stiffness and damping nonlinearity subject to blast loading...........................................................88
Table 5.37: HFTD parameters for the SDOF system with stiffness
and damping nonlinearity subject to blast loading..........................................................................................89
Table 6.1: Numerical data of the onedimensional damreservoir system........................................................103
Table 6.2: Values assigned to the cubic coefficient
Table 7.1: Type of DIANA finite elements used for the case study modeling...................................................119
Table 7.2: Material parameters for the concrete dam....................................................................................120
Table 7.3: Material parameters for the foundation soil..................................................................................120
Table 7.4: Material parameters for the reservoir fluid....................................................................................120
xx
List of Symbols
LIST OF SYMBOLS
xxi
List of Symbols
List of Symbols
Modified pseudo force vector for the time segment k and iteration j
Fluidstructure interaction matrix
Time
Last converged time step
Load signal duration
Time increment or time step size
Traction surface force
Decay function duration
Time segment duration
Displacement of a SDOF system
xxiii
List of Symbols
xxiv
List of Symbols
xxv
List of Symbols
xxvi
1.
Introduction
1. INTRODUCTION
Frequency domain analysis methods are restricted to harmonic loading input. Besides, it is
only possible to obtain the steady state response of a linear elastic system.
Time domain analysis can be applied with any type of loading input. Besides, it is possible to
obtain the transient response of a system including several types of nonlinearities.
However, time domain analysis cannot be applied directly in damreservoir interaction
problems taking into account fluid compressibility or radiation boundary conditions. These
kinds of systems are frequency dependent, and therefore, their equation of motion in the time
domain is defined in terms of convolution integrals. Due to their complexity, these differential
equations with convolution integrals are not directly solved by the classical time domain
methods.
One indirect way of dealing with frequency dependent properties in time domain analysis is
following a weakly coupled approach between the solid and fluid equations of motion, by
combining both of them in a global iterative procedure. Nevertheless, this also means a large
computational effort.
On the other hand, damreservoir interaction problems, including fluid compressibility and
radiation boundary conditions, are naturally solved in the frequency domain by means of a
decoupled solution procedure of the solid and fluid equations of motion. In general, this is the
case for any system with frequency dependent properties.
Furthermore, frequency domain decoupled solution procedure can be applied in combination
with the mode superposition method, attaining a remarkable reduction of the computational
effort.
For these reasons, hybrid methods that combine advantages of both approaches and diminish their
drawbacks are proposed to solve dynamic damreservoir interaction problems, including compressible
fluid and radiation boundary conditions.
In this sense, the research work consists of the study and implementation of the Hybrid Frequency
Time Domain (HFTD) method, applied to the numerical modeling of the linear and nonlinear response
of concrete dams under seismic loading, including dynamic damreservoir interaction, compressible
fluid and radiation boundary conditions.
The thesis includes programming, codification and implementation of the HFTD method in DIANA
finite element software. The transient results obtained with the HFTD method for a realistic model of a
high concrete dam is studied and compared with the results obtained with other time domain analysis
methods available in DIANA.
Through the entire thesis, special attention is given to the study of the following topics:
1.
Introduction
Influence of fluid compressibility and reservoir boundary conditions assumptions (f.e: bottom
absorption) in the seismic response of the system.
Soilstructure interaction between the dam and the rock foundation.
Modeling of the far field nonreflecting boundary conditions for the underground and reservoir
(radiation effect).
Modeling of earthquake forces using base acceleration load (earthquake accelerogram).
Numerical implementation issues of Discrete Fourier Transform (DFT) and HFTD (segmenting
approach, time steps, period duration, pseudo force convergence).
Fluid finite elements.
Interface finite elements for fluidstructure interaction analysis.
What is the influence of the frequency dependent properties (fluid compressibility, radiation
boundary conditions) in the seismic response of damreservoir interaction systems?
Which are the most important numerical implementation issues the HFTD method?
What is the general performance (accuracy, stability and convergence) of the HFTD method
for the seismic analysis of damreservoir interaction problems?
1.
Introduction
Which are the most relevant advantages and drawbacks of the HFTD method for seismic
analysis of damreservoir interaction systems when compared with time domain analysis
methods?
How feasible is the extended and generalized use of the HFTD numerical method in
engineering practice for the computational modeling and seismic analysis of real high dams?
Second, as a sensitivity exercise, the response of a single degree of freedom model of a dam on rigid
foundation is studied. HFTD and Newmark methods are implemented and applied in these models and
the results obtained are compared and interpreted. In the case of the HFTD method, compressibility
and radiation boundary conditions are taken into account in some cases, whereas in others they are
neglected. The effects of these properties in the seismic response of this simple model are analyzed.
Third, the programming, codification and implementation of the HFTD method in DIANA are carried
out. A real scale two dimensional damreservoirfoundation interaction case study is modeled in
DIANA in order to test the HFTD implementation and compare the results obtained with the Newmark
time domain method available in the software.
Finally, conclusions about the influence and importance of compressibility, boundary conditions and
the method of analysis in the seismic response of damreservoir interaction problems are elaborated.
The relevance of HFTD method for this type of problems is evaluated.
The case study is a 100 m. height concrete dam, which foundation material is a flexible soil. In
DIANA, 2D (plane strain) models of this case study are elaborated, putting special attention to the
modeling details and results interpretation. In addition to the seismic time history ground acceleration
loading in the horizontal translational direction, hydrostatic water pressure and gravity loads are
considered as initial conditions.
The contents of the thesis are organized in six chapters. Chapter 2 presents a succinct theory
background review of the frequency domain analysis in single degree of freedom (SDOF) and multi
3
1.
Introduction
degrees of freedom (MDOF) systems. The frequency domain analysis theory is developed using
complex numbers notation. Topics like mode superposition method, direct method and Fourier
analysis are presented.
Chapter 3 develops the Finite Element Method (FEM) formulation of dynamic fluidstructure interaction
(FSI) applied to MDOF damreservoir interaction systems. The discrete fluidstructure coupled
equations of motion are obtained and the influences of the fluid compressibility and the reservoir
boundary conditions are analytically demonstrated. In addition, the modified equation of motion for
base acceleration loading, together with its underlying assumptions and simplifications, are explained.
Chapter 4 develops the HFTD method formulation applied to the MDOF FEM formulation of the damreservoir interaction system presented in Chapter 3. This comprehensive chapter contains the
background theory of the method, and a detailed description of the method procedure,
implementation issues and input parameters.
In Chapter 5 the dynamic response of SDOF systems is studied. For this purpose, the HFTD and
Newmark methods are developed and programmed in MATLAB especially for these SDOF systems.
The MATLAB routines are used to test several types of linear and nonlinear examples, and compare
the results obtained with both methods.
Chapter 6 presents the formulation of a simplified one dimensional damreservoir interaction system
including considerations like fluid compressibility and radiation boundary condition. The frequency
dependent terms introduced by the hydrodynamic pressure are added to the SDOF formulation
developed in Chapter 5. Again, the modified HFTD method is used to test a particular example of a
simplified damreservoir interaction system, and evaluate the influence of compressibility and radiation
boundary in its linear and nonlinear response.
Finally, Chapter 7 presents the results of the HFTD method implementation in DIANA finite element
software. A comprehensive study of the seismic response of 2D foundationdamreservoir interaction
case study modeled with DIANA is performed. The modeling process, as well as the analysis results, is
presented in detail for linear elastic behavior and dams concrete cracking nonlinear behavior. In
addition to the HFTD method, different types of dynamic analysis, like eigenanalysis, frequency
domain analysis (Direct Method) and time domain transient analysis (Newmark method), are also
carried out.
The conclusions and recommendations of the thesis are presented in Chapter 8. Besides, important
information about the HFTD implementation in DIANA, the MATLAB routines developed for SDOF
systems, and the DIANA HFTD command file for the analysis of the 2D foundationdamreservoir
interaction case study, is contained in Appendix A, B and C, respectively.
2.
( )
( )
( )
(2.1)
( )
The solution of Eq. (2.1) (particular solution) is denominated the steady state response of the system,
which in combination with the free vibration solution (complementary solution) and the initial
conditions, provides the total response of the system.
The steady state response of the viscous damped SDOF system defined in Eq. (2.1) is also defined in
the complex plane as expressed in Eq. (2.2). Furthermore, the complex steady state response of the
system can be expressed in terms of the excitation frequency
defined in Eq. (2.3).
(2.2)
( )
( )
(2.3)
Eq. (2.3) is differentiated with respect to time to obtain the expressions for the velocity and
acceleration expressed in Eqs. (2.4) and (2.5), respectively. Both responses can be expressed in terms
of the displacement.
( )
( )
(2.4)
( )
(2.5)
( )
The magnitude of the displacement ( ) is a complex quantity defined in Eq. (2.6) in terms of the real
amplitude ( ) and the phase angle or phase lag ( ) with respect to the excitation force. Different
expressions for the displacement, velocity and acceleration are obtained in Eqs. (2.7) to (2.9),
respectively, after replacing the complex amplitude given by Eq. (2.6) into Eqs. (2.3) to (2.5).
( )
( )
(2.6)
(
(2.7)
)
(2.8)
( )
2.
(2.9)
The systems response and excitation force in the complex plane at time t are illustrated in Fig. 2.1,
which presents a graphical representation of the phase angle between the force and the response, the
magnitude of velocity and acceleration with respect to displacement, and the and rad outofphase
angles of velocity and acceleration, respectively, with respect to displacement.
Figure 2.1: Rotating force, displacement, velocity and acceleration vectors in the complex
plane.
The substitution of Eqs. (2.3) to (2.5) into Eq. (2.1) results in the expression of the displacements
complex amplitude given by Eq. (2.10). From this equation, the expressions for the phase angle
and the real displacement amplitude are determined in Eqs. (2.11) and (2.12).
(2.10)
(2.11)
( )
( )
(2.12)
2.
(2.13)
The extension of Eq. (2.13) to MDOF systems is expressed in Eq. (2.14) in terms of the stiffness
matrix ( ), the damping matrix ( ) and the mass matrix ( ). The displacement amplitude and force
amplitude are defined by the complex vectors and , respectively. Usually, when the excitation
force is defined in the frequency domain, the imaginary part of the force amplitude vector is equal to
zero. However, if the excitation force is defined in the time domain, then a direct Fourier Transform is
required to be expressed in the frequency domain (See Section 2.4), and therefore the force
amplitude vector is complex.
(2.14)
One way of solving Eq. (2.14) is by means of the Direct solution method which converts a n x n
complex system of equations into the 2n x 2n system with real coefficients given by Eq. (2.15). This
doublesized system is obtained as a consequence of splitting the displacement and force amplitude
vectors into their real and imaginary components.
]{
(2.15)
The steady state response of the system, given by the real ( ) and imaginary ( ) parts of the
displacement amplitude vector, is obtained from the solution of Eq. (2.15) for each excitation
frequency .
2
3
2.
vector ( ) is expressed in terms of the complex generalized displacement amplitude ( ) and the mode
shape matrix by means of Eq. (2.17).
(2.16)
(2.17)
Eq. (2.17) is substituted in Eq. (2.14) to obtain the equation of motion in terms of the generalized
system given by Eq. (2.18). The system of equations is reduced to a m x m system (m < n) by
multiplying Eq. (2.18) times the transpose of the mode shape matrix, as expressed in Eq. (2.19) in
terms of the generalized stiffness ( ), mass ( ) and damping ( ) matrices defined in Eqs. (2.20)
to (2.22), respectively.
(2.18)
(2.19)
(2.20)
(2.21)
(2.22)
The generalized displacement vector ( ) is determined from the solution of the reduced equation of
motion defined in Eq. (2.19), which can be solved for each excitation frequency using the direct
method [Eq. (2.15)]. Then is replaced into Eq. (2.17) to obtain .
A more simplified procedure is possible for a special type of damping matrix . Due to the
orthogonality property of the mass and stiffness matrices with respect to the modal shape matrix,
and
or
is proportional
the generalized damping matrix is also diagonal and the system defined by Eq. (2.19) is
uncoupled.
For this particular uncoupled system the solution with the mode superposition method for each
excitation frequency is given by Eqs. (2.23) and (2.24), which are derived from Eqs. (2.17) and
(2.19), respectively.
(2.23)
(2.24)
2.
( )
[ ( )
(2.25)
contained in the loading function ( ) as expressed in Eq. (2.26). The complex magnitude ( ) of
the loading is defined for each harmonic component by the Direct Fourier Transform expressed in Eq.
(2.27).
(2.26)
( )
( )
(2.27)
Furthermore, in order to obtain complex loading magnitude ( ) different than zero when applying
the Direct Fourier Transform defined in Eq. (2.27) ( ) in the denominator tends to infinity), the
change of variable expressed in Eq. (2.31) is introduced.
( )
(2.31)
2.
After replacing Eqs. (2.28) to (2.31) into Eqs. (2.25) and (2.27), and substituting the summatory
operators by integrals, the Direct and Inverse Fourier Transform for nonperiodic loading functions are
defined by Eqs. (2.32) and (2.33), respectively.
( )
( )
( )
( )
(2.32)
)
(2.33)
The continuous time ( ) and circular frequency ( ) variables of Eqs. (2.32) and (2.33) have to be
transformed to their equivalents discrete variables for the DFT formulation. In this sense, the discrete
time points ( ) and frequencies ( ) are defined in Eqs. (2.36) and (2.37), respectively. Similarly, the
differential time ( ) and frequency ( ) is expressed in terms of the time step and frequency interval
by Eqs. (2.38) and (2.39), respectively.
(2.36)
(2.37)
(2.38)
(2.39)
Finally, the substitution of Eqs. (2.34) to (2.39) into Eq. (2.32) and (2.33), and the replacement of the
integrals by summatory operators, provide the expression for the Direct and Inverse DFT defined by
Eqs. (2.40) and (2.41), respectively.
( )
(
[ (
[ ( )
(2.40)
(2.41)
10
Based on Zienkiewicz et al. (2005), Rizos and Karabalis (2000), and Zienkiewicz and Bettes (1978).
11
Therefore, for this type of problems, the linearized NavierStokes dynamic equation of the fluid in
terms of density (), velocity (v), pressure (p) and constant body force of gravity (b), is defined in the
vectorial Eq. (3.1).
(
(3.1)
Considering that a small variation of density is assumed, in Eq. (3.1) can be taken out of the
derivative operator and be replaced by the constant hydrostatic density ( 0). Furthermore, if the body
gravity force is neglected the dynamic equation of fluid is reduced to Eq. (3.2).
(3.2)
On the other hand, the linearized continuity or mass conservation equation of the fluid based on the
same assumptions is given by the scalar Eq. (3.3).
(3.3)
If Eq. (3.4) is replaced into Eq. (3.3) and then it is differentiated with respect to time, the continuity
equation is expressed in terms of pressure by Eq. (3.5).
(3.5)
Eq. (3.2) is differentiated with respect to space and transformed into a scalar equation by applying the
divergence operator. Then it is combined with Eq. (3.5), eliminating the fluid velocity vector variable
(v) and obtaining the classical scalar wave equation given by Eq. (3.6), that governs the fluid domain
motion. The wave speed (c) is defined in terms of the fluid density and bulk modulus, as shown in Eq.
(3.7).
(3.6)
(3.7)
12
Figure 3.1: Fluid and solid domains, fluidstructure interface and boundary conditions
(TNO DIANA BV, 2011).
3.1.3.1.
The expressions describing the boundary conditions should be expressed in terms of pressure (p)
(essential boundary condition) or pressure gradients in the normal direction to the boundary surface
(
Taking this into account, the boundary condition for the fluidstructure interface is obtained by
multiplying the transpose of the normal vector to the boundary surface (nT)4 by Eq. (3.2). Then, the
time derivative of the fluid normal velocity ( ) is prescribed in relation with the pressure gradient in
the normal direction (
(3.8)
(3.9)
Based on Zienkiewicz et al. (2005), Fenves and Chopra (1983), Fenves and Chopra (1984) and TNO DIANA BV
(2011).
3
Generally, for 2D models one of these four boundary conditions belongs to each side of the fluid region.
4
The normal vector n points outward of the fluid domain.
13
The coupling between the fluid and the structure is achieved through the motion continuity of both
physical domains. In this sense, the prescribed fluid normal velocity ( ) is coupled with the structure
normal displacement ( ) as expressed in Eq. (3.10).
(3.10)
The final expression of the fluidstructure interface boundary condition is given by Eq. (3.11), after
replacing Eq. (3.10) into Eq. (3.9).
3.1.3.2.
on
(3.11)
It is possible to specify two types of boundary conditions for the free surface of the reservoir. The first
and simplest one is a consequence of neglecting the effect of the surface waves, prescribing a
pressure equal to zero in the horizontal top free surface, as expressed in Eq. (3.12). This is an
essential or Dirichlet type of boundary condition.
on
(3.12)
The second option for the free surface boundary condition is to use Eq. (3.13) as an approximated
expression to determine the pressure caused by the surface gravity waves. In this case,
is the
vertical elevation (Zdirection) of the fluid relative to the mean surface.
(3.13)
Eq. (3.14) is obtained from Eq. (3.9) after defining z as the normal direction (n).
(3.14)
3.1.3.3.
on
(3.15)
Considering only the fluid wave propagation in Xdirection, the general solution of the wave equation
given by Eq. (3.6) is expressed in Eq. (3.16).
14
(3.16)
On the radiation boundary, which is theoretically located at an infinite distance from the dam in the Xdirection, no incoming waves are entering into the system (only outgoing waves), and therefore Eqs.
(3.17) and (3.18) describe the pressure wave propagation on this boundary.
(3.17)
)
(
(3.18)
Eqs. (3.19) and (3.20) are obtained after differentiate Eq. (3.18) with respect to space (x) and time
(t), respectively.
(3.19)
(3.20)
After eliminating F from Eqs. (3.19) and (3.20), the radiation boundary condition of infinite extent
given by Eq. (3.21) is obtained.
on
3.1.3.4.
(3.21)
Similarly to Eq. (3.6), the wave equation that describes the movement of the bottom materials in
terms of the normal displacement to the bottom surface (
) and wave speed (cr) is defined by Eqs.
(3.22) and (3.23).
(3.22)
(3.23)
Taking into account that in the reservoir bottom only outward propagating waves in the normal
direction are considered, and that there is no incoming propagation waves due to the radiation
condition assumed for the thick layer of bottom materials, the general solution of Eq. (3.22) is given
by Eq. (3.24).
(3.24)
15
On the other hand, the equilibrium equation is obtained in terms of the fluid pressure (defined positive
for compression) and the normal stresses in the bottom materials surface, as expressed by Eq. (3.25).
(3.25)
Eq. (3.26) is obtained after replacing Eq. (3.24) into Eq. (3.25), whereas its derivative with respect to
time is given by Eq. (3.27).
(3.26)
(3.27)
The interaction and continuity between the fluid and the bottom materials is ensured by the
compatibility of displacements, required to derive Eq. (3.28) from Eq. (3.14). Eq. (3.29) is obtained
after replacing Eq. (3.24) into Eq. (3.28).
(3.28)
(3.29)
After combining Eqs. (3.27) and (3.29), the term F is eliminated to obtain Eq. (3.30). Finally, the
bottom absorption boundary condition defined by Eqs. (3.31) and (3.32) is obtained after eliminating
Er through the combination of Eqs. (3.23) and (3.30).
(3.30)
on
(3.31)
(3.32)
To determine the constant q in terms of the wave reflection coefficient ( ), the harmonic pressure
wave propagating in the normal direction of the bottom surface is defined as the summation of a
outward incident wave (pi) and an inward reflected wave (pr), as expressed by Eqs. (3.33) to (3.35).
(3.33)
(
(3.34)
(3.35)
16
After the substitution of Eqs. (3.33) to (3.35) into Eq. (3.31), the relation given by Eq. (3.36) is
obtained for the bottom boundary (n=0). The expression of q in terms of is given by Eq. (3.37),
obtained after the cancellation of the exponential terms of eq. (3.36).
)
(
(3.36)
(3.37)
Eq. (3.37) is replaced into Eq. (3.31) to obtain the bottom absorption boundary condition expressed
by Eq. (3.38) in terms of the wave reflection coefficient ( ).
on
(3.38)
(3.39)
on
(3.40)
The weak form of the solid part of the system is expressed in Eq. (3.41). Besides, Eq. (3.42) is
obtained after applying integration by parts and the Greens theorem over the term containing the
stress tensor (
) in Eq. (3.41).
]
]
(3.41)
(3.42)
The interaction between the fluid and the structure is introduced in the weak form of the solid
equation of motion [Eq. (3.42)] as a Neumann boundary condition in the interface (I), by relating the
surface traction stresses applied in the fluidstructure interface of the solid domain with the
hydrodynamic pressure (p) defined positive in compression. This interaction boundary condition for
the solid domain is expressed in Eq. (3.43).
(3.43)
5
6
After replacing Eq. (3.43) into Eq. (3.42) the final expression of the weak form of the solid part of the
system is obtained in Eq. (3.44). This expression is valid for linear and nonlinear analysis. The most
common nonlinearities, like material or geometric nonlinearities, can be introduced in the terms
and .
(3.44)
With the objective to determine the weak form of the fluid part of the system, its wave equation and
boundary conditions previously defined are reproduced below.
(3.6)
on
(3.11)
on
(3.15)
on
(3.21)
on
(3.38)
The weak form of the fluid part of the system is expressed in Eq. (3.45). Eq. (3.46) is obtained after
applying integration by parts and the Greens theorem over the term
of Eq. (3.45).
]
[
)]
]
(3.45)
[ (
)]
]
(3.46)
(3.47)
(3.48)
After replacing Eqs. (3.47) and (3.48) into Eq. (3.44), the standard Galerkin discretization applied to
the weak form of the solid equation is defined by Eq. (3.49). For the particular case of linear structural
systems, the discrete equation of motion is transformed as shown in Eq. (3.50).
()
( )
( )
(3.49)
(3.50)
The mass (MS), damping (CS), stiffness (KS) and interaction (R) matrices, as well as the external
(
) and internal (
) force vectors appearing in Eqs. (3.49) and (3.50), are directly obtained from
Eqs. (3.44), (3.47) and (3.48). For instance, the expression of the interaction matrix is reproduced
below in Eq. (3.51).
(3.51)
On the other hand, after replacing Eqs. (3.47) and (3.48) into Eq. (3.46), the standard Galerkin
discretization applied to the weak form of the fluid equation is defined by Eq. (3.52). The expressions
defining the matrices MF, CF and KF are given by Eqs. (3.53) to (3.55), respectively.
(3.52)
(3.53)
(3.54)
(3.55)
For this reason, staggered solution procedures are used as a more efficient alternative to solve fluidstructure problems in the time domain. In this type of procedure both equation are solved
independently by means of an artificial uncoupling. This is achieved introducing an assumed or
predicted value of the vector in Eq. (3.49) and solving this equation independently with time
stepping procedures. Consequently, the solution obtained for the vector is then replaced in Eq.
(3.52) and this equation is solved to obtain a new value of . The iterative procedure is repeated until
convergence for both vectors and is achieved.
The advantage of using staggered process is the reduction of the size of the problem and the
computational effort due to the fact that two smaller and uncoupled (instead of one larger and
coupled) systems of equations are solved separately. However, it is important to mention that in most
of the cases staggered solution procedures are only conditionally stable.
In a staggered solution procedure the modified solid domain equation for nonlinear problems is given
by Eq. (3.56) after replacing the predicted nodal pressure vector (
( )
(3.56)
An incremental iterative scheme (f.e. NewtonRaphson method) and a time integration scheme (f.e.
Newmark method) have to be chosen in order to solve Eq. (3.56) and find an approximated solution
for the nodal acceleration vector in the structure (
), which is introduced in Eq. (3.52) to obtain
the modified fluid domain equation for linear behavior given by Eq. (3.57).
(3.57)
Due to linear behavior of the fluid, the solution of Eq. (3.57) only requires the election of a time
integration method, not being required to follow an iterative procedure. A new vector
is
obtained to be introduced again in Eq. (3.56), and the iterative process is continued until convergence
is achieved. Finally, the whole procedure is repeated for the next time step.
]{ }
]{ }
] { }
( )}
(3.58)
It is possible to determine the steady state solution of this coupled system of equations in the
frequency domain by expressing the nodal pressure vector ( ) in terms of the nodal structural
displacement vector ( ). First, the external force vector should be expressed or transformed into the
harmonic form shown in Eq. (3.59).
( )
(3.59)
Considering that both, the structural displacements and the fluid pressures, are produced by the
external force, the steady state solution of both responses will exist in the same harmonic form as
expressed in Eqs. (3.60) and (3.61).
( )
(3.60)
( )
(3.61)
After replacing Eqs. (3.59) to (3.61) into Eq. (3.58), the system of equations shown in Eq. (3.62) is
obtained, where the complex amplitudes of the nodal structural displacements ( ) and pressure ( )
vectors are the unknown variables.
]{ }
(3.62)
From the second row of Eq. (3.62) the amplitude of the nodal pressure vector can be expressed in
terms of the amplitude of the nodal displacement vector as defined in Eqs. (3.63) and (3.64).
(3.63)
( )
(3.64)
After replacing the expression of given by Eq. (3.63) into the first row of Eq. (3.62), the uncoupled
equation to determine the steady state response of the structure in the frequency domain, in terms of
the complex amplitudes of the nodal displacement vector ( ), is given by Eqs. (3.65) to (3.67). The
effect of the reservoir interaction in the dams equation of motion is represented by and , called
added mass and added damping, respectively.
(
(
(3.65)
(3.66)
)
(
(3.67)
Eqs. (3.65) to (3.67) can be solved in the frequency domain using the Direct or the Mode
superposition methods (see Chapter 2).
21
Replacing Eqs. (3.68) and (3.69) into Eq. (3.52) results in the simplified explicit discrete equation
given by Eq. (3.70), which describes the incompressible fluid motion.
(3.70)
In turn, the substitution of Eq. (3.70) into Eq. (3.49) generates a fully uncoupled equation for the
structural part of the system. This simplified structural equation of motion, based on the disregard of
compressible fluid and surface gravity waves, is expressed in Eq. (3.71). It is interesting to note that
the only difference between Eq. (3.71) and the isolated structural system [Eq. (3.49)] is the presence
of the added mass matrix ( ), which expression is given by Eq. (3.72). The simplified Eq. (3.71) for
incompressible fluids can be solved directly in time domain using direct integration or mode
superposition methods, without the necessity of staggered solution procedures.
()
( )
(3.71)
(3.72)
It is not completely clear when it is recommendable to neglect compressibility effects and assume this
simplified special case. One criterion for taking a decision is to estimate the order of magnitude of the
fundamental compressible period of the structure in terms of the dams height (H) and the fluid wave
speed velocity (c), using the expression H/c. If this period is close to the magnitude of the earthquake
acceleration period, then the compressibility effects play an important role and therefore should not
be neglected. This is generally the case for high flexible dams containing a long range of modal
vibration periods. On the other hand, compressibility can be less important in the case of low height
stiffer dams with a compressible fundamental period far smaller than the excitation period.
10
(3.73)
This approach, known as dry dam eigenanalysis, is accurate enough for the calculation of the mode
shape vectors, but not for the calculation of the modal frequencies, which require the consideration of
the fluid effects. One way of doing this is using the full coupled equation defined in Eq. (3.58) to
obtain the free vibration problem expressed in Eq. (3.74).
]{ }
] { }
{ }
(3.74)
However, the matrices of Eq. (3.74) are neither symmetric nor positive definite, which make the
classical computational methods for eigenvalue analysis not directly applicable. Some manipulations
and variable symmetrizing procedures can be followed to solve the eigenvalue problem defined by
Eq. (3.74), yet making the process too complicated.
Another approach known as wet dam eigenanalysis consists in assuming that the interacting fluid is
incompressible. Then, the simplified equation of motion given by Eq. (3.71) can be used to determine
the modal frequencies of the fluidstructure interaction system through the solution of Eq. (3.75). The
influence of the fluid is concentrated in the added mass term ( ) given by Eq. (3.72).
11
(3.75)
Based TNO DIANA BV (2011), Zienkiewicz et al. (2005), Fenves and Chopra (1983) and Darbre (1996).
23
]{
]{
]{
{ }
(3.76)
(3.77)
(3.78)
It can be noticed from Eq. (3.78) that the external force in the absolute equilibrium equation is
produced by the base displacement, velocity and acceleration. However, there are two assumptions
that simplify the force vector given by Eq. (3.78). The first assumption is that in the mass matrix there
is no coupling between the base and superstructure DOF, and therefore Eq. (3.79) holds. The second
assumption, expressed in Eq. (3.80), is to disregard the damping term contribution due to its not
completely clear physical meaning (Wilson, 2002).
(3.79)
(3.80)
12
After replacing the assumptions expressed in Eqs. (3.79) and (3.80) into Eq. (3.78), the force vector
of the absolute equilibrium equation is expressed by Eq. (3.81) in terms of base displacement only.
(3.81)
(3.82)
The matrices ( ) and ( ) represent the unitary rigid body displacements of the superstructure and
the base, respectively, in the three translational directions (X, Y, Z). Eqs. (3.83) and (3.84) show that
each of these rigid body motion matrices is composed by three vectors (one per translational
direction) which values vary between 0 and 1, depending on the cosine direction of each DOF with
respect to the XYZ coordinate system.
( )
( )
( )
( )
( )
( )
( )
(3.83)
( )
(3.84)
Eq. (3.85) shows the superstructure absolute displacement expressed in terms of the rigid body
motion matrix, the ground displacement and the relative displacement ( ). Similarly, Eq. (3.86)
shows a similar expression for the base, with the obvious difference that in this case there is no
relative displacement.
( )
(3.85)
( )
(3.86)
( )
( )
( )
( )
( )
( )
(3.87)
25
Taking into account that internal forces are only proportional to relative displacements (stiffness force)
and velocities (viscous damping force), the multiplication of the stiffness (K) and damping (C)
matrices by the rigid body displacements (I) is equal to zero13, as expressed in Eqs. (3.88) and (3.89).
( )
( )
( )
(3.88)
( )
(3.89)
After inserting Eqs. (3.79), (3.88) and (3.89) into Eq. (3.87), the dynamic equilibrium equation given
by Eq. (3.90) is obtained in terms of the superstructure displacements relative to the base motion
( ). This approximate form of the relative equation of motion is the classic equation used in seismic
analysis.
( )
(3.90)
The base acceleration load can be expressed for each of the three translational directions as shown in
Eq. (3.91), after replacing Eqs. (3.82) and (3.83) into Eq. (3.90).
( )
( )
( )
(3.91)
It is interesting to notice that in the approximate absolute equation of motion [Eqs. (3.77) and
(3.81)], the force vector is proportional to the prescribed base displacements, and it is applied only in
the superstructure nodes close to the base (
). On the other hand, the force vector of the
approximate relative equation of motion [Eq. (3.90)] is proportional to the prescribed base
accelerations, and it is applied in the superstructure nodes not linked with the base (
( )
).
13
For the stiffness matrix this statement is always correct. However, in the case of the damping matrix the statement does not
fulfill if Rayleigh damping definition (aM+bK) is used. The multiplication of the mass proportional term (a0) by the rigid body
motion matrix results in a damping forces different than zero. Nevertheless, this artificial mass proportional damping energy
added to the system is usually ignored (TNO DIANA BV, 2011).
26
4.
4.
With the response in the time domain it is possible to use the constitutive laws ruling the nonlinear
behavior of the system and evaluate the internal forces. Then, a new value of the pseudo forces is
computed as the difference between the linear elastic force and the nonlinear internal force.
Finally, the new value of the pseudo force vector is transformed again to the frequency domain to
solve one more time the linearized equation in the frequency domain. The process is repeated until
convergence is achieved.
4.2. Formulation of the HFTD method for the seismic analysis of a damreservoir
dynamic interaction system2
A general nonlinear equation of motion of the solid part of a fluidstructure interaction problem can
be derived from Eq. (3.49). This equation of motion in absolute coordinates is shown in Eq. (4.1).
(4.1)
( )
For the particular case of earthquake analysis of damreservoir interaction systems, the seismic load
can be applied as a base acceleration loading in the three translational directions, defined by the
ground acceleration vector ( ). This approach is conveniently formulated in relative coordinates
( )
without necessity of the application of an external force (
). Based on Eqs. (3.90) and
(4.1), the nonlinear equation of motion in relative coordinates for the seismic analysis of a damreservoir interaction system is given by Eq. (4.2).
( )
(4.2)
( )
(4.3)
( )
After replacing Eq. (4.3) into Eq. (4.2), and bringing the nonlinear component ( ) to the righthand
side of the equation, a pseudo linear system described by Eq. (4.4) is obtained. The lefthand side of
Eq. (4.4) is completely defined by linear terms, whereas the nonlinearities are concentrated in the
pseudo force vector ( ) located in the righthand side.
( )
( )
( )
(4.4)
Based on Darbre (1996), Fenves and Chavez (1990), Mansur et al. (2000) and Aprile et al. (1994).
28
4.
The most consistent way of taking into account the frequencydependent properties of the reservoirs
compressible fluid, defined by the hydrodynamic force vector (
), is by means of solving the
pseudo linear system described by Eq. (4.4) in the frequency domain. By doing this, in the more
general case, frequencydependent damping can also be included in the system as shown by Eqs.
(4.5) and (4.6).
and
indicate the constant stiffness and damping matrices, respectively,
whereas
( ) is the frequency dependent damping matrix, ( ) is a general frequency dependent
matrix, and
( )
( )
(4.5)
(4.6)
( )
In order to solve Eq. (4.4) in the frequency domain, the excitation force given by the ground
acceleration vector is expressed in the harmonic form shown in Eq. (4.7). Consequently, the harmonic
form of the relative displacement, the hydrodynamic pressure and the pseudo force vectors are
expressed in Eqs. (4.8) to (4.10), respectively.
( )
( )
(4.7)
(4.8)
( )
(4.9)
( )
( )
(4.10)
( )
After replacing Eqs. (4.7) to (4.10) into Eq. (4.4), the exponential common terms are eliminated, and
as a result, the equation of motion of the damreservoir system in the frequency domain is defined by
Eq. (4.11) in terms of the complexvalued amplitudes of the ground acceleration [ ( )], the relative
dams displacement [( )], the hydrodynamic pressure [ ( )] and the pseudo force vector [( )].
] ( )
( )
( )
( )
( )
(4.11)
Based on Eq. (3.52), the complexvalued frequency dependent expression of the hydrodynamic
pressure amplitude is determined from the fluid equation of motion given by Eq. (4.12).
(4.12)
Eq. (4.12) is transformed into Eq. (4.13) after expressing the dam absolute acceleration vector ( ) in
terms of the ground ( ) and relative ( ) acceleration vectors.
( )
(4.13)
29
4.
After replacing Eqs. (4.7) to (4.9) into Eq. (4.13), the complexvalued frequency dependent
expression of the hydrodynamic pressure amplitude is defined by Eqs. (4.14) and (4.15).
( )
( )
( )
( )
(4.14)
( )]
(4.15)
Eq. (4.16) is obtained by inserting the expression of the hydrodynamic pressure amplitude given by
Eq. (4.14) into the equation of motion of the coupled damreservoir system defined by Eq. (4.11). The
hydrodynamic mass matrix ( ) that appears in both sides of Eq. (4.16) is a complexvalued
frequencydependent term defined by Eq. (4.17), which represents the added mass effect in the
dam provided by the reservoir.
] ( )
( )
( )
(4.16)
( )
(4.17)
( )
Another possibility of expressing the effect of the reservoir on the dam is to split
in its real (added
mass) and imaginary (added damping) components, as shown in Eq. (4.18). The expressions for
the added mass ( ) and damping ( ) are given by Eqs. (4.19) and (4.20), respectively.
] ( )
( )
( )
( )
(4.18)
(4.19)
)
(
(4.20)
The mode superposition method can be employed by introducing the variable defined by Eq. (4.21).
Consequently, the premultiplication of Eq. (4.18) by the matrix of modal vectors transposed (
)
results in the equation of motion, defined by Eqs. (4.22) to (4.28), in terms of the generalized modal
coordinates vector ( ).
( )
[
(4.21)
( )
( )
] ( )
( )
( )
(4.22)
(4.23)
( )
( )
(4.24)
(4.25)
30
( )
4.
(4.26)
(4.27)
( )
(4.28)
The equation of motion in the generalized modal coordinates constitutes a condensed system of
equations in which the unknowns are reduced to a number of chosen modes of vibration. In the
general case only the frequency independent matrices
and
are diagonal, which means that the
system is not uncoupled.
One way of uncouple numerically Eq. (4.22) is to express the full matrices
,
and
as the
summation of a diagonal (d) and an offdiagonal (od) terms as shown in Eqs. (4.29) to (4.31).
(4.29)
(4.30)
(4.31)
Replacing Eqs. (4.29) to (4.31) into Eq. (4.22), and bringing the offdiagonal terms to the righthand
side, results in an uncoupled system of equations defined by Eq. (4.32) which matrices in the lefthand side are diagonal. The uncoupled system can be solved iteratively by treating the offdiagonal
terms as a second set of pseudo forces dependent of the generalized modal coordinates vector and
defined by Eq. (4.33).
( )
( )
( )
( )
( )
(4.32)
( )
(4.33)
Nevertheless, the introduction of additional pseudoforce vectors to impose the uncoupling of the
system of equations could lead to a cumbersome numerical process, especially in large models with
high amount of DOF.
31
4.
Based on Darbre and Wolf (1988), Fenves and Chavez (1990), Darbre (1990), Chavez and Fenves (1993),
Chavez and Fenves (1994) and Veletsos and Ventura (1985).
32
4.
START
)] and
( )
( )
()
( )
Solve the pseudo linear Eq. (4.22) and determine the generalized
response in the frequency domain ( ), ( ), ( )
NO
Convergence?
YES
FINISH
33
4.
For a detailed discussion of HFTD method stability criteria see Darbre and Wolf (1988).
34
4.
The Fourier period (Tp) should be defined as the summation of the ground acceleration duration (t o)
plus an additional extended period called quiet zone (Tq). The quiet zone is a band of zeros added at
the end of the ground acceleration signal, which length depends on the free vibration properties of
the system (fundamental period of vibration and viscous damping ratio of the structure).
Steps 6, 8 and 10 of the HFTD solution procedure outlined in Section 4.3.1 indicate that the Fourier
transform is applied to the earthquake ground acceleration (direct), the pseudo force (direct) and the
response of the system (inverse), respectively. All these transformations should always be performed
over the same Fourier period (Tp), which is calculated using Eqs. (4.36) and (4.37), in terms of the
fundamental vibration period ( ) and the viscous damping ratio of the system ( ) (Chavez and
Fenves, 1984).
(4.36)
(4.37)
Eq. (4.37) shows that for long oscillation periods (low frequency systems) and low damping ratio, the
quiet zone (Tq) is longer. In the extreme case of a damping ratio equal to zero Tq tends to infinity,
which means that the HFTD method is not applicable for undamped systems (Veletsos and Ventura,
1985).
As mentioned before, the quiet zone is necessary to satisfy the initial conditions and consequently
obtain a better approximation of the transient response instead of the steady state response. On the
contrary, the use of Fourier transformation period which is too short provides inaccurate results, no
matter that they be stable and apparently reasonable (Darbre and Wolf, 1988). In conclusion, the
insertion of a quiet zone should be understood as a numerical representation of what theoretically
should be an infinity period in the analytic Fourier integral (See Section 2.4).
Fig. 4.2 shows a ground acceleration signal which Fourier Transform is calculated twice, one using a
Fourier period equal to the ground acceleration duration (
), and the other using an
extended Fourier period that includes the quiet zone (
).
35
4.
The spectrums of the ground acceleration amplitudes obtained in both cases are shown in Figs. 4.3
and 4.4, respectively. It can be noticed that both spectrums have similar shapes, with the difference
that the one obtained with the extended Fourier period (Fig. 4.4) is smoother due the higher number
of points used in its definition.
The extended period is ten times the loading duration, and therefore, the number of points used in
both cases holds the same relation. This is the reason why the magnitudes of the ground acceleration
amplitudes of Fig. 4.4 are around ten times the amplitudes of Fig. 4.3.
In order to obtain the transient response of the system, the HFTD method uses the ground
acceleration spectrum shown in Fig. 4.4.
Figure 4.3: Spectrum of the ground acceleration amplitude for Fourier period equal to the
loading duration (
)
Figure 4.4: Spectrum of the ground acceleration amplitude for extended Fourier period
including quiet zone (
)
36
method
4.
solution
procedure
using
the
time
Using a time segmentation approach implies the repetition of the HFTD solution outlined in Section
4.3.1 (from step 6 to step 15) for all the segments defined in the time span of interest. It is important
to realize that the time span of interest (T r) is not the same as the Fourier period (Tp), but the time
for which the calculation of the system response is desired.
Logically, the time span of interest (Tr) is shorter than the period used for the Fourier transform (T p).
As shown in Eq. (4.38), the minimum value of Tr includes the ground acceleration duration (t o) plus
and additional free vibration time equal to one half of the fundamental period of vibration, during
which a maximum peak response could still be attained. The time segmentation should be performed
only over Tr, and the HFTD solution procedure is concluded when the last segment of T r has
converged.
(4.38)
Fig. 4.5 shows the kth time segment and the time steps contained inside. The initial points of time
segments k (Tk) and k+1 (Tk+1), the last converged time step (tc) of the previous iteration, the
time step size (t), and the time segment size (Ts) can also be identified from Fig. 4.5. The initial
time step of segment k is calculated with Eq. (4.39).
(4.39)
The HFTD procedure outlined in Section 4.3.1 indicates that, for the first iteration of the first time
segment only, the pseudo force vector is set equal to zero (step 5) and transformed to the frequency
domain (step 8) using the Fourier transformation period (Tp). Once the pseudo linear system has been
solved in the frequency domain (step 9), the generalized response is transformed back to the time
domain (step 10).
37
4.
For the subsequent iterations, the segmentation approach requires that the pseudo forces and
displacements in the time domain are calculated and updated (step 11 to step 14) only for the not
converged time steps belonging to the current time segment (tc < t Tk+1  t). The iterative process
(step 7 to step 14) is repeated until the convergence criteria (step 15) are fulfilled for all the time
steps inside the current segment of analysis (tc = Tk+1  t).
When all the time steps inside the current segment have converged (tc = Tk+1  t), the next segment
is investigated. However, in the first iteration of the new segment (step 6) the values of the pseudo
forces belonging to the previous converged time segments are maintained (the pseudo force values of
the converged previous time steps are not modified nor updated anymore), whereas for the current
and subsequent segments, the values of the pseudo forces are set to zero (the definition of a
decaying appended function is also possible as explained in the next section).
In this way, the iterative procedure is continued until the last segment of the time span of interest is
covered, doing always the Fourier transformation through all the period T p, but updating progressively
the displacement and pseudo force vectors only for the nonconverged time steps inside the current
segment of analysis. The modified version of the HFTD method flowchart, including time
segmentation approach in the solution procedure, is shown in Fig. 4.6
As a final point, special attention should be given to the fact that the equation of motion of the
pseudo linear system (not the real nonlinear system) is the one solved in the frequency domain.
Therefore, the eigenvalues, eigenvectors and critical damping ratios, required to perform the mode
superposition method and to determine the length of the Fourier transformation period (T p) [Eqs.
(4.36) and (4.37)], should be based on the properties of the pseudo linear system defined in Eq.
(4.22).
38
4.
START
)] and
( )
( )
()
( )
Solve the pseudo linear Eq. (4.22) and determine the generalized
response in the frequency domain ( ), ( ), ( )
NO
Convergence?
YES
YES
Figure 4.6: Flowchart of the modified HFTD method solution procedure including the time
segmentation approach
39
4.
4.3.5. Definition of the appended decaying functions for the ground acceleration
loading and the pseudo force
Due to the segmentation approach for solving the linearized equation in the frequency domain, both
load vectors in the righthand side of Eq. (4.22), ground acceleration [ ( )] and pseudo force
[ ( )], are filled with zeros for all the time steps belonging to the time segments coming after the
current. Therefore, the expressions defining ( ) and ( ), for the time segment k and iteration
j, are given by Eqs. (4.40) and (4.41), respectively, in terms of the Heaviside function (H).
( )
( )[ ( )
( )
(4.40)
)]
( )[ ( )
)]
(4.41)
This means that both time load vectors die out suddenly between the time steps T k+1 t and
Tk+1. In order to avoid this sudden unloading and consequent inaccurate results, a decaying function
may be appended to both load components. The decaying function must start in the last point of the
current time segment (k+1 t) and its extension (Td) is recommended not to be longer than one
segment length (Ts), due to implementation more than performance or accuracy reasons (Darbre,
1990).
One type of decaying function that can be used for this purpose is the sinusoidal function proposed by
Darbre (1990), which is defined by Eqs. (4.42) and (4.43). This decaying function matches the value,
slope and curvature of the loading signal in the time step k+1 t.
( )
[ ( )
( )
( )] [
(4.42)
(4.43)
As expressed in Eq. (4.40), the ground acceleration signal is taken from 0 to Tk+1 t, neglecting
the subsequent part. Instead, the decaying function [Eq. (4.42)] is appended to the signal in the time
point Tk+1 t, followed by a band of zeros until the Fourier period (Tp) is completed.
Therefore, the modified ground acceleration signal for the time segment k is expressed in Eq.
(4.44), where the definition of
is given by Eq. (4.45). Fig. 4.7 is a graphical representation of Eq.
(4.44) applied to the ground acceleration loading shown in Fig. 4.2.
( )
( )[ ( )
)]
( )
(4.44)
(4.45)
40
4.
Figure 4.7: Modified excitation load with appended decaying function for the HFTD
analysis of the kth time segment.
On the other hand, for the first iteration the pseudo force vector is set equal to zero for all the time
steps between 0 and Tp. For the next iterations, the values of the pseudo force vector corresponding
to the converged time steps (from 0 to tc) are kept fixed, whereas for the other not converged time
steps inside the current segment (from tc to Tk+1 t), the pseudo force values corresponding to the
last iteration are assigned.
Similarly as in the case of the earthquake ground acceleration, the pseudo force values of the
following time segments correspond to the appended decaying function, continuing with a band of
zeros until completing the Fourier period (Tp). The modified pseudo force vector for the time segment
k and iteration j is expressed in Eq. (4.46).
( )
( )[ ( )
)]
( )
(4.46)
( )

( )
( )
( )

( )
( )
(4.47)
(4.48)
41
4.
(4.50)
(4.51)
The value of the maximum sampling frequency (F) is inversely proportional to the time step size (t).
This means that once t is defined, F is determined using Eq. (4.53).
(4.53)
Based on Chavez and Fenves (1994), Veletsos and Ventura (1985), Craig and Kurdila (2006) and TNO DIANA
BV (2011).
6
In case of analysis including damfoundation interaction, the mass of the soil should not be taken into account
in the calculation of the effective mass.
42
4.
(4.54)
When the problem to be solved with the HFTD method is highly nonlinear (f.e. problems including
contact or base sliding nonlinearities), it is recommended that in addition the time step size fulfills Eq.
(4.55) (Chavez and Fenves, 1994).
(4.55)
The number of time point samples (N) given by Eq. (4.57) is obtained from the Fourier period (Tp)
and the time step size (t). The frequency increment ( ) is calculated using Eq. (4.58), whereas the
maximum sampling frequency (F) is calculated with Eq. (4.59), which is equivalent to Eq. (4.53).
(4.57)
43
4.
(4.58)
(4.59)
44
5.
5.1 Solution procedure of the HFTD method for SDOF nonlinear systems
The HFTD method, which was previously explained in detail for seismic analysis of damreservoir
dynamic interaction multi degree of freedom (MDOF) systems (Chapter 4), is developed here for the
particular case of SDOF dynamic linear and nonlinear systems.
It was previously stated that the HFTD method was developed to analyze systems with frequency
dependent properties and nonlinear behavior. Therefore, the formulation of the HFTD method for
SDOF systems also considers the inclusion of frequency dependent terms, no matter that in most of
the SDOF numerical examples developed in Section 5.3, these frequency dependent systems are not
taken into account. This is because to evaluate the soundness, precision and stability of the HFTD
method, the results obtained for the SDOF systems have to be compared with other analytical
solutions or established numerical methods (f.e.: Newmark method) in the time domain, which are
naturally capable of dealing with nonlinearities, but not with frequency dependence.
The SDOF system considered is expressed by Eq. (5.1). As it can be noticed, this equation of motion
constitutes a general case with nonlinear behavior [
] and frequency dependent terms
[
]. Therefore, it is not possible to directly solve Eq. (5.1) in the time domain.
(5.1)
Assuming that additive decomposition between linear and nonlinear terms is possible, the internal
nonlinear restoring force is expressed by Eq. (5.2) in terms of the pseudo force [
], the constant
stiffness ( ) and the constant damping ( ) (Aprile et al., 1994).
(5.2)
After replacing Eq. (5.2) into Eq. (5.1) and passing the pseudo force to the right hand side, Eq. (5.3)
is obtained. This equation of motion contains only linear terms in the left hand side, whereas all
sources of nonlinearities are concentrated in the pseudo force term located in the right hand side of
Eq. (5.3).
(5.3)
Obviously, the value of the pseudo force is not known a priori, and for that reason Eq. (5.3) (known
as pseudo linear or linearized equation) has to be solved iteratively until convergence is achieved for
both, the displacement solution ( ) and the pseudo force ( ). Based on Eq. (5.2), Eq. (5.4) shows the
expression to calculate the value of the pseudo force in iteration j.
(5.4)
45
5.
If there were no frequency dependent terms, the pseudo linear equation could be solved in both,
frequency and time domains. However, frequency dependent terms are considered in Eq. (5.3) and
therefore, the HFTD method solves the pseudo linear system in the frequency domain.
Accordingly, the excitation force is expressed in terms of its harmonic components defined in Eq.
(5.5). As a consequence, the displacement and the pseudo force are defined in the same terms by
Eqs. (5.6) and (5.7), respectively. N is the number of excitation frequencies ( ) contained in the
force function (or signal)
.
(5.5)
(5.6)
(5.7)
Eq. (5.8) is the pseudo linear equation of motion in the frequency domain, which is solved for each
excitation frequency ( ). Eq. (5.8) is obtained in terms of the complex valued amplitudes of the
excitation force [
], displacement [
] and pseudo force [
], after replacing Eqs. (5.5)
to (5.7) into Eq. (5.3).
(5.8)
Prior to solve Eq. (5.8), the complex right hand side amplitudes [
] have to be
determined. Then, the equation is solved and the complex amplitude of the displacement [
] is
obtained. The HFTD method uses the Discrete Fourier Transform (DFT) defined in Eqs. (5.9) and
(5.10) to determine
and
, respectively.
(5.9)
(5.10)
The number of excitation frequencies (N) considered in the analysis depends on the period used for
the Fourier transform (Tp) and the time step size (t) that determines the separation of the discrete
time points in which the response of the system is computed. Eq. (5.11) presents the expression to
determine N, whereas Eq. (5.12) the expression to calculate
.
(5.11)
(5.12)
46
5.
In Chapter 4 it was explained that in the HFTD method the Fourier period (T p) is not equal to but
longer than the duration of the excitation force (t0). The reason for this is to obtain the transient
response of the system instead of the steady state response (Section 4.3.3). The expression to
calculate the ideal value of the Fourier period is given by Eq. (5.13), where the natural period of the
undamped system (T1) and the viscous damping ratio ( ) are defined by Eqs. (5.14) and (5.15),
respectively. For the SDOF systems treated in this chapter, the very long value of Tp obtained with Eq.
(5.13) is totally manageable in terms of computational time and memory capacity.
(5.14)
(5.13)
(5.15)
The fundaments and procedure of the time segmentation approach followed by the HFTD method to
achieve convergence when solving nonlinear systems have already been explained in Section 4.3. This
procedure consists in dividing the time span of interest (not the total Fourier period) in time segments
over which the iterative solution is found sequentially for each of these segments.
The time span of interest (Tr) is the time for which the user desires to obtain the response of the
system. Usually, Tr is equal or longer than the load duration but much smaller than the Fourier period
(t0 Tr << Tp).
On the other hand, all the time segments are generally composed of a constant number of time steps.
Therefore, the size of each time segment (TS) is calculated with Eq. (5.16), in terms of the number
of samples (or time points) per segment (N S) and the time step size (t), whereas the starting time
point of segment k (Tk) is calculated using Eq. (5.17).
(5.16)
(5.17)
5.
Figure 5.1: Excitation load, segmentation approach, time span of interest and Fourier
period.
Figure 5.2: Modified excitation load with appended decaying function for the HFTD
analysis of the kth time segment.
The use of a decaying function avoids a sudden unloading of the system, and consequent inaccuracy
and artificial oscillations in the obtained response. Taking into account that in the pseudo linear
equation of motion solved by the HFTD method, the frequency dependent complex amplitude terms
located in the right hand side of Eq. (5.8) do not only consist in the Fourier transform of the load
48
5.
[
], but of the pseudo force [
] as well, a similar modification shown in Fig. 5.2 for
be accomplished for
before the DFT defined in Eq. (5.10) is applied.
has to
The modifications of the excitation load and the pseudo force for the analysis of the kth time segment
are analytically expressed in Eqs. (5.18) and (5.19), respectively, by means of the Heaviside function
[
].
(5.18)
(5.19)
].
Nevertheless, the simplest decaying function defined by Eqs. (5.20) and (5.21) is selected for both of
them. This is a linear decaying function defined in terms of its time variable ( ) and its duration
(
).
starts in the last point of the time segment k, whereas by definition
should be smaller
than the time segment length (
).
(5.20)
(5.21)
Going back to the solution of Eq. (5.8), the complex amplitude of the displacement ( ) is determined
in the frequency domain for each excitation frequency ( ). The complex amplitudes of the systems
velocity and acceleration in the frequency domain are determined for each excitation frequency using
Eqs. (5.22) and (5.23).
(5.22)
(5.23)
From the amplitudes of the system response in the frequency domain, the response in the time
domain is obtained by means of the inverse DFT, expressed in Eqs. (5.24) to (5.26) for the
displacement, velocity and acceleration, respectively.
(5.24)
(5.25)
(5.26)
49
5.
Due to the segmentation approach followed in the HFTD method, the response in the time domain is
only required to be calculated for those time points located inside the current segment being
analyzed, for which convergence has not been achieved yet. This means that if the last time point in
the segment k for which convergence has been achieved in the previous iteration is denominated ,
then the time points
, for which Eqs. (5.24) to (5.26) have to be evaluated, fulfill the condition
expressed in Eq. (5.27).
(5.27)
after replacing
and
in Eq. (5.4). Then, the convergence criteria check is performed for
both, the pseudo force and the displacement. The convergence criteria verification for iteration j is
presented in Eqs. (5.28) and (5.29).


(5.28)
(5.29)
50
5.
Table 5.1: Summary of the HFTD method for SDOF nonlinear systems
1.
Initial calculations
1.1
1.2
1.3
1.4
2.
Fourier period
2.1
2.2
3.
4.
3.2
4.2
4.3
4.4
5.
Set up the initial the value of the pseudo force equal to zero
6.
6.2
6.3
]
[
Follow the HFTD iterative procedure to determine the system response and pseudo forces of the
time points of segment k (See Table 5.2)
51
5.
Table 5.2: Iterative procedure of the HFTD method for SDOF nonlinear systems
1.
Determine the modified pseudo force for time segment k and iteration j1 [
1.2
1.3
Determine the modified pseudo force for time segment k and iteration j1 [
1.4
1.5
Solve the pseudo linear equation to determine displacement amplitude in the frequency
domain
[
1.5.2
] and acceleration [
domain
1.6
1.7
1.8
between
1.9


1.10
Update
1.11
between
and
and
)
between
and


5.
domain. This permits to evaluate the HFTD method performance and generate confidence about the
results obtained when frequency dependent properties are actually required to be included.
Nevertheless, it is obvious that the numerical examples tested in the section 5.3 cannot have
frequency dependent properties, due to fact that the benchmark solutions are obtained from time
domain analysis methods.
This section is divided in two parts. The first part covers the analytic solution in the time domain of
SDOF linear systems. Two particular types of loading are considered: harmonic and general. The
second part develops the wellestablished Newmark method in the time domain for SDOF nonlinear
systems.
5.2.1 Analytic solution in the time domain for SDOF linear systems1
The equation of motion of the damped SDOF linear system shown in Fig. 5.3 is expressed by Eq.
(5.30) in terms of the relative displacement ( ). The initial conditions expressed in Eqs. (5.31) and
(5.32) complete the definition of the governing equations.
(5.30)
(5.31)
Based on Craig and Kurdila (2006) and Paz and Leigh (2004).
(5.32)
53
5.
Eq. (5.30) can be expressed as shown in Eq. (5.33). The definition of the natural frequency of
vibration ( ) and the viscous damping ratio ( ) is presented in Eqs. (5.34) and (5.15), respectively.
(5.33)
(5.34)
omitted, a general equation valid for relative and absolute coordinates is expressed in Eq. (5.35).
(5.35)
5.2.1.1
When the excitation force acting in the righthand side of Eq. (5.35) is harmonic in the form
, then Eq. (5.35) can be expressed as shown in Eq. (5.36).
(5.36)
The solution of Eq. (5.7) is given by the summation of the general homogeneous solution and the
particular solution. For underdamped systems (
), the general solutions is defined by Eqs. (5.37)
and (5.38).
(5.37)
(5.38)
On the other hand, the particular solution or steady state response is given by Eq. (5.39) in terms of
the displacement amplitude ( ) and the phase angle ( ), defined by Eqs. (5.40) and (5.41),
respectively.
(5.39)
(5.40)
) ]
)]
(5.41)
54
5.
Therefore, the total solution of the SDOF underdamped linear system subjected to the harmonic load
is given by the sum of Eqs. (5.37) and (5.39), as expressed in Eq. (5.42). The derivative
of Eq. (5.42) provides the expression of the velocity of the system, shown in Eq. (5.43). The
acceleration can be obtained from the equation of motion, after the substitution of Eqs. (5.42) and
(5.43) into Eq. (5.36). The values of
and
are obtained from the initial conditions defined in Eqs.
(5.31) and (5.32).
(5.42)
(5.43)
5.2.1.2
Response of SDOF systems subjected to general loading by the
Convolution Integral method
If the loading is not harmonic but a general dynamic excitation
, the particular solution is
obtained by the Convolution (or Duhamel) integral method, which is expressed in Eq. (5.44). The total
solution is given therefore by Eq. (5.45).
(5.44)
(5.45)
5.2.1.3
Response of SDOF systems subjected to general loading by the
Direct Integration method
In many applications, the excitation force
cannot be expressed by a mathematical function (f.e:
ground acceleration signal records) and therefore, the convolution integral of Eq. (5.45) cannot be
obtained in a continuous closed form.
One way to deal with these kind of problems without losing accuracy and theoretical consistency is to
evaluate the convolution integral in a discrete form by dividing the loading in small linear segments, as
shown in Fig. 5.4.
In this way, if the time step size (t) that determine the position of the discrete points in the force
function is small enough, the solution of the system can be calculated for the time step i, using as
initial conditions the solution obtained for the time step i1. This approach is known as the Direct
Integration Method and provides the exact analytical solution of the system, provided that the time
step size is small enough to describe accurately the force function.
55
5.
The force for the time interval between ti and ti+1 is approximated to a linear function by Eqs. (5.46)
and (5.47). Therefore, the equation of motion (only valid for the time interval between t i and ti+1) is
expressed in Eq. (5.48).
(5.46)
( )
(5.47)
[(
( )
(5.48)
The general and particular solutions of the system for the interval i are given by Eqs. (5.49) and
(5.50), respectively. The constants Ci and Di of the general solution are obtained based on the initial
conditions of the interval (i.e.
and ). On the other hand, the constants Ai and Bi are obtained
after replacing the particular solution in the equation of motion defined in Eq. (5.48).
(5.49)
(5.50)
The total solution for displacement and velocity is given respectively by Eqs. (5.51) and (5.52),
respectively.
(5.51)
(5.52)
56
5.
After replacing Eq. (5.50) into Eq. (5.48), the expressions defined in Eqs. (5.53) and (5.54) are the
solution for Ai and Bi, respectively.
(5.53)
(5.54)
(5.56)
Finally, the solution of interval i is completely defined and therefore it is possible to determine the
response of the next time step i+1, by replacing the time
in Eqs. (5.51) and (5.52), to obtain
and
respectively. The acceleration (
) is obtained from Eq. (5.48).
After substitution and simplification, the following final recurrence formulas for displacement, velocity
and acceleration of the time step i+1 are presented in Eqs. (5.57) to (5.59). The expressions to
calculate the coefficients of the recurrence formulas are summarized in Table 5.3.
(5.57)
(5.58)
(5.59)
57
5.
Table 5.3: Coefficients for the recurrence formulas of the Direct Integration Method
[(
5.2.2 Solution procedure of the Newmark method for SDOF nonlinear systems2
The equation of motion for nonlinear systems solved by the HFTD method [Eq. (5.1)] is modified in
such a way that no frequency dependent terms are taken into account, as it is expressed in Eq.
(5.60). This general equation for nonlinear SDOF systems is valid for absolute and relative
coordinates, and therefore, the force vector
can include a ground acceleration excitation.
(5.60)
One of the most stable and accurate time domain methods for nonlinear analysis of structural
dynamics systems is the Newmark method. Due to implementation purposes of Newmark method, it is
desirable to split the nonlinear internal force [
] in stiffness [
] and damping
[
] nonlinear forces. This is expressed in Eq. (5.61).
(5.61)
Eq. (5.61) can be discretized so it is valid at discrete time instants t i. Therefore, the nonlinear equation
of motion is satisfied in the time instants ti and ti+1, as it is expressed in Eqs. (5.62) and (5.63).
(5.62)
(5.63)
58
5.
The difference between Eqs. (5.63) and (5.62) gives the incremental equilibrium equation expressed
in Eq. (5.64), for which the time step is defined by Eq. (5.65)
(5.64)
(5.65)
The incremental stiffness and damping internal restoring forces can be expressed as shown in Eqs.
(5.66) and (5.67), respectively. This approximation is valid for small step sizes (t) and for tangent
stiffness and damping coefficients defined by Eqs. (5.68) and (5.69), respectively.
(5.66)
(5.67)
(5.68)
(5.69)
After replacing Eqs. (5.66) and (5.67) into Eq. (5.64), the incremental equilibrium equation is
reformulated and presented in Eq. (5.70). This equation has the same form of the equation for linear
systems, and this is the reason for splitting the restoring force in Eq. (5.61).
(5.70)
Now, the Newmark method can be applied in the same fashion as it is done for linear systems. The
only differences are that the tangent stiffness and damping have to be updated for each time step,
and that an iterative procedure has to be included to avoid the error accumulation of pure incremental
procedures (modified NewtonRaphson iterative method)3.
The Newmark method approximates the displacement and velocity of the system in the time instant
ti+1 via the recurrence algorithms presented in Eqs. (5.71) and (5.72), respectively. The values of the
parameters and determine the stability and accuracy of the method. The two most used cases due
to their high accuracy are the linear (=1/2, =1/6) and average (=1/2, =1/4) acceleration
methods.
(5.71)
(5.72)
See section 5.7 of Chopra (2007), for a detailed explanation of the modified NewtonRaphson iterative
procedure applied to nonlinear dynamic analysis of SDOF systems.
59
5.
After rearranging Eqs. (5.71) and (5.72), the incremental velocity and acceleration are expressed in
Eqs. (5.73) and (5.74), respectively, in terms of the incremental displacement and the response of the
time instant ti.
(5.73)
(5.74)
Eqs. (5.73) and (5.74) are substituted into Eq. (5.70) to obtain the incremental equilibrium equation,
defined by Eqs. (5.75) to (5.77). The only unknown of Eq. (5.75) is the incremental displacement
(
), which is calculated based on the solution of the previous time instant t i, and following the
modified NewtonRaphson iterative procedure in order to obtain accurate results.
(5.75)
(5.76)
(5.77)
The incremental displacement found after solving Eq. (5.75) can be replaced into Eq. (5.73) to find
. Then, the displacement, velocity and acceleration of the time instant t i+1 are calculated using
Eqs. (5.78) to (5.80), respectively.
(5.78)
(5.79)
(5.80)
The internal restoring forces in the time instant t i+1 are calculated incrementally, as it is expressed in
Eqs. (5.81) and (5.82). The definition of the restoring forces increment is given by Eqs. (5.66) and
(5.67).
(5.81)
(5.82)
Table 5.4 shows a schematic summary of the Newmark method for SDOF nonlinear systems, whereas
Table 5.5 shows the procedure of the NewtonRaphson iterative procedure.
60
5.
Table 5.4: Summary of the Newmark method for SDOF nonlinear systems4
Special cases:
(1) Average acceleration method (
Initial calculations
1.1
1.2
2.
2.2
2.3
2.4
2.5
2.6
2.7
) and velocity (
;
2.8
;
2.9
61
5.
Initial calculations
1.1
1.2
1.3
) and damping (
2.
iteration j1
2.2
2.3
Determine the total displacement increment of iteration j between time steps i and i+1 (
2.4
Determine the total velocity increment of iteration j between time steps i and i+1 (
2.5
2.6
Determine the internal restoring forces of iteration j for time step i+1 (
) based on
{
2.8
2.9
)
)
62
5.
5.3.1.1
Harmonic loading6
Table 5.6: Numerical data of Example 4.2 (Craig and Kurdila, 2006)
k
40
m
0.1
0.2
p0
10
10
t0
2
Based on the information given in Table 5.6, the parameters of the SDOF system shown in Table 5.7
are determined. The analytical solution for harmonic loading developed in section 2.1 is used to solve
this example. Table 5.8 shows the displacement amplitude (U), phase angle ( ) and constants A1 and
A2, which are obtained using Eqs. (5.40) to (5.43), as well as the information shown in Tables 5.6 and
5.7.
Table 5.7: Parameters of the SDOF linear system subject to harmonic loading
Tr
2
T1
0.314
20
19.60
Table 5.8: Analytic solution of SDOF linear system subject to harmonic loading
U
0.322
0.261
0.311
0.1059
Finally, Eqs. (5.83) and (5.84) present the analytic displacement and velocity, obtained from Eqs.
(5.42) and (5.43), respectively. The acceleration response is indirectly determined by replacing the
displacement and velocity values into Eq. (5.36).
(5.83)
5
6
63
5.
(5.84)
On the other hand, Table 5.9 shows the values of the parameters required to perform the HFTD
method schematized in Tables 5.1 and 5.2. It is important to mention that for linear systems the
HFTD method does not require segmentation approach or iterative procedure. For that reason only
one time segment (NT =1) is used in all the examples corresponding to linear systems.
Table 5.9: HFTD parameters for the SDOF linear system subject to harmonic loading
t
Tp
0.02
15.72
786
NT
0.0636
The comparison between the analytic and the HFTD solutions for the displacement, velocity and
acceleration is presented in Figs. 5.5 to 5.7.
For the displacement and velocity responses, the major differences of the HFTD solution with respect
to the analytical solution are found during the first 0.5 s. For the acceleration, this range is extended
until the time 0.75 s. After this initial period, the HFTD response gets stable and matches almost
exactly the analytical solution thenceforth.
The reason for this initial inaccurate range is that the magnitude of the cosine harmonic load does not
start in zero, but in p0 equal to 10. The HFTD is not capable to exactly trace the response generated
by this sudden loading at the initial time t=0. Once the effect of the sudden initial loading is
diminished, the response obtained with the HFTD method gradually matches the analytical solution.
Figure 5.5: Harmonic loading example. Displacement response for Analytic and HFTD
solutions.
64
5.
Figure 5.6: Harmonic loading example. Velocity response for Analytic and HFTD solutions.
Figure 5.7: Harmonic loading example. Acceleration response for Analytic and HFTD
solutions.
5.3.1.2
Blast loading7
The SDOF system is subject to the blast load schematized in Fig. 5.8. The numerical data defining the
characteristics of the system is shown in Table 5.10.
7
65
5.
Table 5.10: Numerical data of Example 4.2 (Paz and Leigh, 2004)
k
100 000
m
100
0.2
t0
0.06
Based on the information given in Table 5.10, the parameters of the SDOF system shown in Table
5.11 are determined. The recurrence formulas of the Direct Integration method developed in Section
5.2.1.3 are used to calculate the exact analytic response of the SDOF linear system. Table 5.12 shows
the values of the coefficients for the recurrence formulas of the Direct Integration method. They were
calculated based on the information given in Tables 5.10 and 5.11, as well as the expressions of Table
5.3.
Table 5.11: Parameters of the SDOF linear system subject to blast loading
Tr
0.20
T1
0.20
31.62
30.92
Table 5.12: Direct Integration method coefficients for the SDOF linear system subject to
blast loading
t
0.005
6.32
A
8.1x108
B
4.1x108
C
0.9878
D
0.0048
A
2.4x105
B
2.4x105
C
4.825
D
0.9268
On the other hand, Table 5.13 shows the values of the parameters required to perform the HFTD
method (Tables 5.1 and 5.2). Similarly as in the previous example, for linear analysis the
segmentation approach is not required (only one time segment is required to be defined).
Table 5.13: HFTD parameters for the SDOF linear system subject to blast loading
t
Tp
0.005
9.935
1987
NT
0.1007
1
66
5.
The comparison between the analytic Direct Integration and the HFTD solutions for the displacement,
velocity and acceleration is presented in Figs. 5.9 to 5.11. The accuracy obtained in the three
responses is remarkable. The variations are so small that are almost no perceptible. This means that
when the loading starts in zero (see Fig. 5.8) the HFTD method shows a great performance.
Figure 5.9: Blast loading example. Displacement response for Direct Integration and HFTD
solutions.
Figure 5.10: Blast loading example. Velocity response for Direct Integration and HFTD
solutions.
67
5.
Figure 5.11: Blast loading example. Acceleration response for Direct Integration and
HFTD solutions.
5.3.1.3
Impulsive loading8
The SDOF system is subject to the impulse load schematized in Fig. 5.12. The numerical data defining
the characteristics of the system is shown in Table 5.14.
68
5.
Table 5.14: Numerical data of Example 4.7 (Paz and Leigh, 2004)
k
10 000
m
10
0.1
t0
0.10
Based on the information given in Table 5.14, the parameters of the SDOF system shown in Table
5.15 are determined. Table 5.16 shows the values of the coefficients for the recurrence formulas of
the Direct Integration method [Eqs. (5.57) to (5.59)].
Table 5.15: Parameters of the SDOF linear system subject to impulsive loading
Tr
0.22
T1
0.20
31.62
31.46
Table 5.16: Direct Integration method coefficients for the SDOF linear system subject to
impulsive loading
t
0.01
3.16
A
3.2x106
B
1.6x106
C
0.9514
D
0.0095
A
4.7x104
B
4.9x104
C
9.53
D
0.8912
On the other hand, Table 5.17 shows the values of the parameters required to perform the HFTD
method. Similarly as in the previous example, for linear analysis only one time segment is required.
Table 5.17: HFTD parameters for the SDOF linear system subject to impulsive loading
t
Tp
0.01
9.94
994
NT
0.1006
The comparison between the analytic Direct Integration and the HFTD solutions for the displacement,
velocity and acceleration is presented in Figs. 5.13 to 5.15. The accuracy obtained in the three
responses is as good as in the previous example.
69
5.
Figure 5.13: Impulsive loading example. Displacement response for Direct Integration
and HFTD solutions.
Figure 5.14: Impulsive loading example. Velocity response for Direct Integration and
HFTD solutions.
70
5.
Figure 5.15: Impulsive loading example. Acceleration response for Direct Integration and
HFTD solutions.
5.3.1.4
The SDOF system is subject to the ground acceleration correspondent to the NorthSouth component
of the El Centro earthquake shown in Fig. 5.16. The numerical data defining the characteristics of the
system is shown in Table 5.18.
71
5.
Table 5.18: Numerical data of Example 4.9 (Paz and Leigh, 2004)
K
t0
2136
38.86
0.05
4.5
Based on the information given in Table 5.18, the parameters of the SDOF system shown in Table
5.19 are determined. Table 5.20 shows the values of the coefficients for the recurrence formulas of
the Direct Integration method. On the other hand, Table 5.21 shows the values of the parameters
required to perform the HFTD method.
Table 5.19: Parameters of the SDOF linear system subject to earthquake ground
acceleration
Tr
4.38
T1
0.85
7.414
7.405
Table 5.20: Direct Integration method coefficients for the SDOF linear system subject to
earthquake ground acceleration
t
0.02
0.37
A
3.4x106
B
1.7x106
C
0.9891
D
0.0198
A
2.5x104
B
2.5x104
C
1.09
D
0.9744
Table 5.21: HFTD parameters for the SDOF linear system subject to earthquake ground
acceleration
t
Tp
0.02
42.38
2119
NT
0.0236
The comparison between the analytic Direct Integration and the HFTD solutions for the displacement,
velocity and acceleration is presented in Figs. 5.17 to 5.19. The accuracy obtained in the three
responses is as good as in the previous example.
72
5.
Figure 5.17: Ground acceleration loading example. Displacement response for Direct
Integration and HFTD solutions.
Figure 5.18: Ground acceleration loading example. Velocity response for Direct
Integration and HFTD solutions.
73
5.
Figure 5.19: Ground acceleration loading example. Acceleration response for Direct
Integration and HFTD solutions.
5.3.2 Resonance case study: SDOF linear systems subject to harmonic loading
with exponential amplitude
The SDOF linear system is subject to a harmonic sinusoidal load which amplitude varies exponentially,
as expressed by Eq. (5.85). In order to test the stability of the HFTD method, the influence of load
excitation frequencies close to the resonance condition ( = o) is studied for systems with low
damping ratios ( 0.05)10. In addition, the loading duration (t0) is expressed in Eq. (5.86) as a
number n of load excitation periods (T) or peaks.
(5.85)
(5.86)
The numerical data of the SDOF system to be studied is shown in Table 5.22. As it can be noticed,
three different low damping ratios are taken into account.
Table 5.22: Numerical data of the SDOF system for resonance case studies
k
2136
m
38.86
7.414
First, four load excitation frequencies close to the natural vibration frequency of the system are
considered with a damping ratio equal to 0.05. The characteristics of these four case studies are
summarized in Table 5.23.
10
As explained in Section 4.3, the HFTD method is not applicable to undamped systems.
74
5.
Table 5.23: Numerical data of case study A: Excitation frequencies close to resonance
Case
A1
A2
A3
A4
0.05
0.05
0.05
0.05
0.80
0.95
1.05
1.20
1.06
0.89
0.81
0.71
2500
2500
2500
2500
25
25
25
25
26.49
22.31
20.19
17.65
Figure 5.20: Harmonic sinusoidal load with exponential amplitude for the case study A4.
Like in the previous examples, the results obtained with the HFTD method (Tables 5.1 and 5.2) are
compared with the analytical solution provided by the Direct Integration method described in Section
5.2.1.3, and summarized by Eqs. (5.57) to (5.59) and Table 5.3.
The HFTD parameters for each of the four case studies are summarized in Table 5.24. Due to lack of
space, only the results obtained for the displacement response of the SDOF linear system are
presented and compared with the Direct Integration method. The analysis of velocity and acceleration
results revealed as good performance as the displacement results, presented in Figs. 5.21 to 5.24 for
the cases A1, A2, A3 and A4, respectively.
Table 5.24: HFTD parameters for case study A
Case
Tp
A1
A2
A3
A4
0.025
0.025
0.025
0.025
53.00
44.68
42.38
42.38
2120
1787
1695
1695
NT
0.0189
0.0224
0.0236
0.0236
1
1
1
1
75
5.
Figure 5.21: Resonance case study A1. Displacement response for Direct Integration and
HFTD solutions.
Figure 5.22: Resonance case study A2. Displacement response for Direct Integration and
HFTD solutions.
76
5.
Figure 5.23: Resonance case study A3. Displacement response for Direct Integration and
HFTD solutions.
Figure 5.24: Resonance case study A4. Displacement response for Direct Integration and
HFTD solutions.
From the analysis of Figs. 5.21 to 5.24, the first conclusion is that the accuracy of the HFTD method is
not affected due to load excitation frequencies close to natural vibration frequency of the system
(resonance condition).
Second, there is a remarkable similarity between the responses of the case studies with excitation
frequencies which are symmetric with respect to the resonance frequency. This is the situation of case
studies A1 and A4 on the one hand, and case studies A2 and A3 on the other hand, which time
history displacements have similar forms and amplitudes. The only small difference, besides the
obvious fact that the responses of the cases with higher load excitation frequencies have a higher
77
5.
number of peaks too, is that the amplitudes of the cases with excitation frequencies below resonance
are slightly greater than its correspondent symmetric cases above resonance.
Finally, it is remarkable the increase of the displacement magnitude for the case studies closer to the
resonance frequency. The maximum displacement magnitude increases from 3 (A1, A4) to 8 (A2, A3), approximately.
Then, based on the good performance shown by the HFTD method in the previous case study, more
critical situations of very lightly damped SDOF linear systems close to resonance condition are
considered. Table 5.25 shows three additional case studies with load excitation frequencies equal to
the resonance frequency and three low damping ratios equal or smaller than 0.05.
Table 5.25: Numerical data of case study B: Resonance excitation frequency with very low
damping ratios
Case
B1
B2
B3
0.05
0.03
0.01
0.85
0.85
0.85
2500
2500
2500
25
25
25
21.19
21.19
21.19
The HFTD parameters for each of the three case studies are summarized in Table 5.26. Similarly than
in case study A, Figs. 5.25 to 5.27 only present the results obtained for displacements, which are
compared with the Direct Integration method response. The analysis of velocity and acceleration
results revealed as good performance as displacement results.
Table 5.26: HFTD parameters for case study B
Case
Tp
B1
B2
B3
0.025
0.025
0.025
42.43
42.43
84.78
1697
1697
3391
NT
0.0236
0.0236
0.0118
1
1
1
Figure 5.25: Resonance case study B1. Displacement response for Direct Integration and
HFTD solutions.
78
5.
Figure 5.26: Resonance case study B2. Displacement response for Direct Integration and
HFTD solutions.
Figure 5.27: Resonance case study B3. Displacement response for Direct Integration and
HFTD solutions.
From the analysis of Figs. 5.25 to 5.27, it is clear that the HFTD method remains accurate with very
low damped systems in resonance conditions. In addition, the comparison between Fig. 5.25 ( =
0.05) and Fig. 5.27 ( = 0.01) shows the notorious effect of damping in the amplitude and stability of
motion. The displacement of case study B3 (system with extremely low damping) growths in time
without bound, until reaching a maximum amplitude equivalent to two and four times the maximum
amplitude of case studies B2 and B1, respectively.
79
5.
5.3.3.1
Linear elastoplastic nonlinear stiffness behavior subject to a half
cycle sine pulse11
The
SDOF
system
described
in
Table
5.27
is
subject
to
the
half
cycle
sine
pulse
[
] of the system is shown in Fig. 5.29. Based on the information given in Table 5.27, the
parameters of the SDOF system shown in Table 5.28 are determined.
Table 5.27: Numerical data of Example 5.5 (Chopra, 2007)
k
10
m
0.2533
0.05
Rt
7.5
Rc
7.5
Table 5.28: Parameters of the SDOF system with stiffness nonlinearity subject to sine
pulse loading
Tr
0.90
T1
1.0
6.283
6.275
11
80
5.
Newmark and HFTD methods are applied in the system using two different time steps in order to
evaluate the improvement of the HFTD accuracy, when reducing the time step size. For the Newmark
method, the Average Acceleration special case was performed, according to the procedure outlined in
Tables 5.4 and 5.5. The parameters required for the execution of Newmark method are shown in
Table 5.29.
Table 5.29: Newmark method (average acceleration) parameters for the SDOF system
with stiffness nonlinearity subject to sine pulse loading
t1
0.1
t2
0.01
On the other hand, Table 5.30 shows the values of the parameters required to perform the HFTD
method, outlined in Tables 5.1 and 5.2, for each of the two time step sizes taken into account. In this
case more than one time segment (NT) is required in order to achieve convergence. The number of
time segments required increases for smaller time steps due to the higher amount of time points and
excitation frequencies (N) included in the analysis.
Table 5.30: HFTD parameters for the SDOF system with stiffness nonlinearity subject to
sine pulse loading
t
Tp
0.1
0.01
50.00
50.00
500
5000
NT
0.02
0.02
10
13
81
5.
The comparison between the Newmark time domain method and the HFTD method solutions for the
displacement, velocity and acceleration is presented in Figs. 5.30 to 5.32. In addition, the nonlinear
hysteretic relation between the stiffness force and the displacement is presented in Fig. 5.33.
The most remarkable observation is that all the responses obtained using both methods are almost
identical when a time step (t) equal to 0.01 s is used. On the other hand, when the time step is
equal to 0.1 s, the accuracy of both methods diminishes, being possible to identify small drifts with
respect to the more accurate solution (t = 0.01 s). However this is a shortcoming of the selected
time step size, and not of the particular characteristics of the methods applied. This can be clearly
appreciated in Fig. 5.33, where the nonlinear hysteretic elastoplastic relation between the stiffness
force and the displacement is shown. When an abrupt change of the tangent stiffness is produced, the
proper description of the nonlinear relation is only possible when small time steps are used.
Figure 5.30: Linear elastoplastic stiffness force nonlinearity. Displacement response for
Newmark and HFTD solutions.
Figure 5.31: Linear elastoplastic stiffness force nonlinearity. Velocity response for
Newmark and HFTD solutions.
82
5.
Figure 5.32: Linear elastoplastic stiffness force nonlinearity. Acceleration response for
Newmark and HFTD solutions.
Figure 5.33: Linear elastoplastic stiffness force nonlinearity. Hysteretic stiffness forcedisplacement relation for Newmark and HFTD solutions.
83
5.
5.3.3.2
Cubic elastoplastic nonlinear stiffness behavior subject to a blast
loading12
The SDOF system described in Table 5.31 is subjected to the blast loading shown in Fig. 5.34.
Table 5.31: Numerical data of Example 8.15 (Humar, 2002)
k
8
m
0.10
0.1118
Rt
8
Rc
8
The nonlinear stiffness diminishes from 8 to 0, which is the point where the stiffness force reaches its
maximum tensile strength equal to 8. After this, the plastic part of the nonlinear relation starts, the
displacement increases and the velocity diminishes until zero. The negative velocity makes the system
going back in the direction of the equilibrium position following a straight line path with a slope equal
to the initial stiffness (see Fig. 5.35). After the stiffness force is reduced to zero, the system enters in
the compression zone under the same cubic elastic relation given by Eq. (5.87). The hysteretic cycle is
repeated until the whole time span of interest is covered.
12
84
5.
Based on the information given in Table 5.31, the parameters of the SDOF system shown in Table
5.32 are determined.
Table 5.32: Parameters of the SDOF system with stiffness nonlinearity subject to blast
loading
Tr
0.90
T1
0.702
8.950
8.894
Two different time steps are also used in this test example. For the Newmark method, the Average
Acceleration special case was performed using the parameters shown in Table 5.33. On the other
hand, Table 5.34 shows the values of the parameters required to perform the HFTD method for each
of the two time step sizes taken into account.
Table 5.33: Newmark method (average acceleration) parameters for the SDOF system
with stiffness nonlinearity subject to blast loading
t1
0.1
t2
0.01
Table 5.34: HFTD parameters for the SDOF system with stiffness nonlinearity subject to
blast loading
t
Tp
0.1
0.01
35.2
35.13
352
3513
NT
0.0284
0.0285
9
9
85
5.
The comparison between the solutions of the Newmark and the HFTD methods for the displacement,
velocity and acceleration responses is presented in Figs. 5.36 to 5.38. In addition, the nonlinear
hysteretic relation between the stiffness force and the displacement is presented in Fig. 5.39.
Figure 5.36: Cubic elastoplastic stiffness force nonlinearity. Displacement response for
Newmark and HFTD solutions.
Figure 5.37: Cubic elastoplastic stiffness force nonlinearity. Velocity response for
Newmark and HFTD solutions.
86
5.
Figure 5.38: Cubic elastoplastic stiffness force nonlinearity. Acceleration response for
Newmark and HFTD solutions.
Figure 5.39: Cubic elastoplastic stiffness force nonlinearity. Hysteretic stiffness forcedisplacement relation for Newmark and HFTD solutions.
From the analysis of Figs. 5.36 to 5.39, it is clear that using a time step equal to 0.1 s does not
generate accurate results. The accuracy of the HFTD method is more affected by the time step size
due to the relation that this parameter has with the range of frequencies included in the Fourier
analysis. When the time step is large, the maximum excitation frequency included in the HFTD
analysis is small, and as a consequence, the results obtained are less accurate.
However, similarly than in the previous example, a sufficiently small time step (t = 0.01 s) ensures
the high precision of the results obtained, not only for the HFTD method, but also for the time domain
Newmark method.
87
5.
Finally, Fig. 5.39 shows that the use of a time step equal to 0.01 s makes it possible to trace the cubic
elastic stiffness forcedisplacement relation with good precision, which is not possible when using a
larger time step. Furthermore, both Newmark and HFTD methods show that the last two time points
calculated are located in the cubic elastic compression zone. This information is not identified with the
larger time step.
5.3.3.3
Cubic elastoplastic nonlinear stiffness and displacement dependent
nonlinear damping behavior subject to a blast loading
The SDOF system is exactly the same of the previous example (Table 5.31) with the only difference
that the damping force is neither linear nor exclusively proportional to velocity, but also displacement
dependent, as expressed in Eq. (5.89). The initial tangent damping is defined by Eq. (5.90), whereas
the value of the constant damping coefficient (c) is calculated with Eq. (5.15). The blast loading used
in this example is also the same as the previous example (Fig. 5.34).
(5.89)
(5.90)
Eq. (5.90) shows that the initial tangent damping can be positive or negative. Both cases are
calculated using Newmark and HFTD methods, but considering just one time step equal to 0.01 s, for
which it has been shown in the previous examples that the more accurate results are obtained. The
values of the parameters c and for the definition of the damping force in Eq. (5.89) are shown in
Table 5.35.
Table 5.35: Parameters for the definition of the nonlinear damping force
c
0.20
t
0.01
88
5.
Table 5.37: HFTD parameters for the SDOF system with stiffness and damping
nonlinearity subject to blast loading
t
Tp
0.01
35.13
3513
NT
0.0285
It is important to state that for both cases (negative and positive initial tangent damping) studied in
this example, the parameters of both methods are the same. What is more, the pseudo linear
equation in the frequency domain of the HFTD method (step 1.5.1 of Table 5.2) is always solved using
a positive value of c, no matter the sign of the initial tangent damping given in Eq. (5.90). Therefore,
the difference between one case and the other is concentrated in the expression for calculating the
nonlinear internal damping force in step 1.7 of Table 5.2 (HFTD) and step 2.6 of Table 5.5
(Newmark).
For both cases (positive and negative initial tangent damping), the comparison between the Newmark
time domain method and the HFTD method solutions is presented in Figs. 5.40 to 5.42, for the
displacement, velocity and acceleration, respectively. As in the previous cases, it is shown that using
an adequate time step size (t = 0.01) makes the HFTD method as precise as the Newmark method,
even for this type of highly nonlinear systems defined by Eqs. (5.87) and (5.89).
Moreover, the nonlinear hysteretic relations stiffness forcedisplacement and damping forcevelocity
are presented in Figs. 5.43 and 5.44. Both figures confirm that the precision of the HFTD method is
not affected by the type (stiffness or damping nonlinearity) and degree (quadratic or cubic) of nonlinearity, neither by the sign of the initial tangent stiffness or damping.
89
5.
90
5.
5.
response (previously shown) for the displacement, velocity and acceleration in Figs. 5.45 to 5.47,
respectively. In this section only the HFTD solutions are displayed, taking into account that in the
previous section it was demonstrated that HFTD provides the same results as Newmark method.
In general terms, Fig. 5.45 shows that nonlinear behavior increases the magnitude of the
displacement response. The effect of damping nonlinearity described by Eq. (5.89) depends on the
sign of the initial tangent damping. If it is positive, the damping force increases since the beginning
and therefore the magnitude of the displacement response is reduced. On the other hand, if the initial
damping is negative, the initial damping force has the same direction of the movement and the
displacement magnitude is amplified.
Figs. 5.46 and 5.47 show that for velocity and acceleration responses, the first positive peak value is
higher for damping nonlinearity with negative initial tangent damping. Nevertheless, once the effect
of the initial damping is vanished, the linear elastic behavior provides the higher peaks.
Figure 5.45: Effect of nonlinear behavior. Displacement response for HFTD solution.
Figure 5.46: Effect of nonlinear behavior. Velocity response for HFTD solution.
92
5.
Figure 5.47: Effect of nonlinear behavior. Acceleration response for HFTD solution.
93
5.
94
The structural part of such a onedimensional model can be represented by a one DOF massspringdashpot system which reflects its inertial, stiffness and damping properties. On the other hand, the
fluid part can be modeled as a rod which longitudinal motion is described by the onedimensional
wave equation, the boundary conditions in both ends of the rod, and the physical and geometric
properties of the fluid domain.
This simplified onedimensional model of a damreservoir interaction system excited by a ground
acceleration loading is shown in Fig. 6.2.
The system shown in Fig. 6.2 is solved using two approaches. The first one is the displacement
approach in the time domain (Section 6.2), whereas the second one is the hydrodynamic pressure in
the frequency domain (Section 6.3), which follows the same formulation explained in Chapter 3 for 3D
MDOF damreservoir interaction systems.
In Section 6.4 some damreservoir interaction examples are modeled and analyzed with both
approaches. Due to the fact that the accuracy of HFTD method for SDOF systems is proven in Chapter
5, all the simulations developed in Section 6.4 are carried out using this method.
Finally, the HFTD equation of motion of the SDOF model shown in Fig. 6.3 is given by Eq. (6.2) in
terms of the relative displacement amplitude of the dam ( ), the ground displacement amplitude
( ), the pseudo force ( ) containing the systems nonlinearity, and the nonreflective viscous
damping (
] ( )
) ( )
( )
(6.2)
It is important to notice that in Eq. (6.2) the damping force generated by the nonreflective boundary is
proportional to the absolute velocity, whereas the damping force of the structure is proportional only to the
velocity relative to the ground.
96
(6.3)
In order to solve this equation and find the analytic expression of the pressure wave propagation, two
boundary conditions (one per each end of the rod) have to be defined. The first boundary condition is
given by the interaction between the structural mass and the fluid, whereas the second one is given
by the hypothesis of the reservoir length (finite or infinite) extension.
The pressure wave equation approach to model the onedimensional damreservoir interaction system
is graphically described in Fig. 6.4. Depending on the assumption for the far extent boundary
condition two different cases can be developed (Sections 6.3.1 and 6.3.2), whereas, if the fluid is
assumed to be incompressible, a third case (Section 6.3.3) is obtained for this approach.
97
In any of the three particular cases the interaction between the fluid and the structure is reduced to a
frequency dependent pressure or surface load [ ( )] applied to the structural massspringdashpot
system2, as it is shown in Fig. 6.5. For this reason, the onedimensional model under the pressure
approach is ideal to be solved with the HFTD method.
The HFTD equation of motion of the SDOF model shown in Fig. 6.5 is given by Eq. (6.4) in terms of
the relative displacement amplitude of the dam ( ), the ground displacement amplitude ( ), the
pseudo force ( ) containing the systems nonlinearity, and the hydrodynamic pressure load ( ) due
to the interaction with the reservoir.
] ( )
( )
( )
( )
(6.4)
The analytic expressions of ( ) for each of the three particular cases previously mentioned are
derived in the next sections.
on
(6.5)
The second boundary condition is expressed by Eq. (6.6) which defines the radiation boundary at
infinite extent (
). This boundary condition, known as the Sommerfelds radiation condition, was
previously presented in Eq. (3.21).
on
(6.6)
The HFTD method solves in the frequency domain the system shown in Fig. 6.4. Consequently, it is
convenient to express the excitation force (ground acceleration) and the response (relative
displacement) of the system in their harmonic form, as shown in Eqs. (6.7) and (6.8), respectively.
( )
(6.7)
( )
(6.8)
In the same way, the pressure wave generated in the fluid domain due to the excitation force in the
structural domain is given by the summation of the incident and reflected harmonic waves. Eq. (6.9)
expresses both harmonic waves in terms of their frequency dependent complex amplitudes and the
wave velocity of propagation.
( )
( )
(6.9)
The expression of the harmonic wave defined in Eq. (6.9) fulfills the wave equation of motion defined
in Eq. (6.3). Therefore, in order to calculate the unknown amplitudes of the incident and reflected
waves, it is required to use both boundary conditions previously defined. After replacing Eq. (6.9) into
the infinite extent radiation boundary condition given by Eq. (6.6), the relations expressed in Eqs.
(6.10) and (6.11) are obtained.
Notice that in Eq. (6.5) the total acceleration is expressed as the sum of ground and relative acceleration.
99
( )
(6.10)
(6.11)
From Eq. (6.10) it can be concluded that the amplitude of the reflected pressure is equal to zero
[ ( )
], which is expected due to the nonreflective nature of Sommerfelds radiation boundary
condition. Therefore, the pressure wave is only composed by the incident wave, and Eq. (6.9) is
reformulated as expressed in Eq. (6.12).
( )
(6.12)
After replacing Eqs. (6.7), (6.8) and (6.12) into the second boundary condition at
given by Eq.
(6.5), the expression of the incident pressure wave amplitude is obtained in Eq. (6.13). The frequency
dependent amplitude of the hydrodynamic force [ ( )] is obtained in Eq. (6.14) after multiplying the
pressure times the interaction surface area ( ).
( )
[ ( )
( )]
( )
[ ( )
( )]
(6.13)
(6.14)
The HFTD equation for the particular Case I is obtained in Eq. (6.15) by replacing the hydrodynamic
force amplitude given by Eq. (6.14) into Eq. (6.4). It is important to notice that Eq. (6.15) is identical
to Eq. (6.2). This is because that, in spite of being derived by different formulations, both equations
responds to the same hypothesis for the fluid domain, namely, compressible with radiation boundary
of infinite extent.
] ( )
) ( )
( )
(6.15)
Eq. (6.15) shows that for Case I the reservoir interaction with the structure provides additional
damping and frequency dependent loading terms in the left and righthand sides of the equation,
respectively.
6.3.2. Case II: Compressible fluid with zero pressure boundary of finite extent
The zero pressure boundary condition of finite extent is defined in Eq. (6.16) for a reservoir with finite
length equal to
. After replacing Eq. (6.9) into Eq. (6.16) the relation between the incident and
reflected pressure amplitudes is obtained in Eq. (6.17).
on
(6.16)
100
(6.17)
The substitution of Eqs. (6.7) to (6.9) and Eq. (6.17) into the fluidstructure interface boundary
condition, given by Eq. (6.5), provides the expression of the incident pressure amplitude in Eq. (6.18).
The reflected pressure amplitude is obtained in Eq. (6.19) after replacing Eq. (6.17) into Eq. (6.18).
( )
( )
[ ( )
[ ( )
(6.18)
( )]
(6.19)
( )]
The summation of both amplitudes given by Eq. (6.20) provides the total amplitude of the pressure
wave acting at the fluidstructure interface. The frequency dependent hydrodynamic force amplitude
[ ( )] is obtained in Eq. (6.21) after multiplying the pressure times the interaction surface area ( )
and simplify the Euler notation complex numbers division.
( )
( )
)
(
[ ( )
[ ( )
(6.20)
( )]
(6.21)
( )]
The HFTD equation for the particular Case II is obtained in Eq. (6.22) by replacing the hydrodynamic
force amplitude given by Eq. (6.21) into Eq. (6.4). Due to the finite extent reflexive boundary
condition, the fluid interaction with the structure provides exclusively frequency dependent additional
mass to the left and righthand side of Eq. (6.22).
)
(
] ( )
)
(
) ( )
( )
(6.22)
(6.24)
101
(6.25)
Due to the lack of fluids compressibility, radiation boundary condition of infinite extent is not possible
to be considered. Therefore, Eqs. (6.24) and (6.25) are introduced in the boundary condition of finite
extent at
, given by Eq. (6.16). The relation between the constants
and
is obtained in
Eq. (6.26).
(6.26)
Replacing Eqs. (6.7), (6.8), (6.24) and (6.25) into the fluidstructure interface boundary condition
given by Eq. (6.5), and then using the relation of Eq. (6.26), provide the expressions of
and
,
given by Eqs. (6.27) and (6.28), respectively.
[ ( )
(6.27)
( )]
[ ( )
(6.28)
( )]
After replacing Eqs. (6.27) and (6.28) into Eq. (6.25), the space and frequency dependent amplitude
of the hydrodynamic pressure is obtained in Eq. (6.29). However, it is of interest to know what is the
pressure amplitude at
, where the mass of the structure is located. This pressure acting in the
fluidstructure interface [ ( )], as well as its associated force [ ( )], is defined in Eqs. (6.30) and
(6.31), respectively.
(
( )
)[ ( )
(
[ ( )
[ ( )
( )
( )]
(6.29)
(6.30)
( )]
(6.31)
( )]
The HFTD equation for the particular Case III is obtained in Eq. (6.32) by replacing the hydrodynamic
force given by Eq. (6.31) into Eq. (6.4). Eq. (6.32) shows that the hydrodynamic force for the
incompressible fluid case is an additional inertial force provided by the mass of the fluid body (
).
This means that under this assumption, the mass of the fluid behaves like a rigid body oscillating
attached to the structure and introducing no damping to the system, like in the previous particular
cases.
] ( )
) ( )
( )
(6.32)
102
) is calculated with Eq. (6.35), and the mass of water contained in the reservoir
The ground acceleration loading applied to the base of the dam corresponds to the first 4.5 s of the
El Centro earthquake NorthSouth component shown in Fig. 6.6.
Table 6.1: Numerical data of the onedimensional damreservoir system
Dam
(N/m)
6.6 x 10
0.05
Reservoir
(kg)
3.6 x 107
(m)
2500
(m)
200
(kg/m3)
1000
(m/s)
1440
(6.33)
(6.34)
(6.35)
(6.36)
104
Figure 6.7: Damreservoir interaction SDOF systems. Displacement response for linearelastic behavior.
Figure 6.8: Damreservoir interaction SDOF systems. Velocity response for linearelastic
behavior.
105
Figure 6.9: Damreservoir interaction SDOF systems. Acceleration response for linearelastic behavior.
Figure 6.10: Damreservoir interaction SDOF systems. Base reaction response for linearelastic behavior.
(6.37)
106
As explained before, Cases I and II are frequency dependent systems which take into account the
reservoir compressibility. Therefore, it is of interest to study the effect of the nonlinear behavior in
these two frequency dependent systems. Additionally, the nonlinear response of the dry dam case
(no frequencydependence) is also presented as a reference.
For the three cases the initial tangent stiffness is assumed to be equal to the linear elastic stiffness of
the dam ( ), as expressed in Eq. (6.38). However, the values of the cubic coefficient ( ) in Eq.
(6.37) are assigned for each case as shown in Table 6.2.
(6.38)
( )
The previous analysis of Figs. 6.7 to 6.10 demonstrates that the reflective boundary condition (Case
II) generates higher linearelastic responses, and therefore a stiffer and more resistant nonlinear
behavior is required for this particular system. This is the reason why a smaller value of
is assigned
to this case in Table 6.2.
Finally, the pseudo force in the time domain for the cubic stiffness softening nonlinearity is calculated
using Eq. (6.39). Then, the nonlinear behavior is completely defined and the HFTD method can be
applied in the solution of the SDOF damreservoir interaction models.
( )
(6.39)
107
First, the nonlinear behavior of the dry dam case is studied as a reference of a non frequency
dependent system. Figs. 6.12 and 6.13 present the effect of nonlinear behavior in the displacement
and stiffness internal force of the SDOF system, respectively.
Figure 6.12: Dry dam case. Effect of nonlinearity in the displacement response.
Figure 6.13: Dry dam case. Effect of nonlinearity in the stiffness force response.
Fig 6.12 shows that the major effects of nonlinear behavior occur between 2 s and 4 s. As expected,
the system including softening nonlinearity generates the higher displacement peak value after the
time 2 s. Fig 6.13 shows that this is the most faraway point located in the plastic regime (horizontal
branch), where the displacement of the system increases with no additional resisting stiffness force.
After the time 4.5 s the ground acceleration loading ends and the system comes back in free vibration
to the linear elastic regime, where both linear and nonlinear responses dampout with a small phase
and the same magnitude.
108
Second, the nonlinear behavior of the frequency dependent Case I, representing a damreservoir
interaction system with a radiation boundary of infinite extent, is studied. Figs. 6.14 and 6.15 present
the effect of nonlinear behavior in the displacement and stiffness internal force, respectively, of this
SDOF frequency dependent system.
Figure 6.14: Damreservoir interaction with infinite extent radiation boundary (Case I).
Effect of nonlinearity in the displacement response.
Figure 6.15: Damreservoir interaction with infinite extent radiation boundary (Case I).
Effect of nonlinearity in the stiffness force response.
Due to the additional damping introduced to the system by the infinite extent boundary condition [Eq.
(6.15)], the influence of nonlinear behavior is less important in this case than in the dry dam case.
Certainly, Fig. 6.14 shows that there are increments of the displacement response peak values around
time 2 s, however these are not as significant as in Fig. 6.12. The same can be concluded from Fig.
109
6.15 which shows that the displacements produced are not large enough to reach the horizontal
yielding branch and develop a more important nonlinear behavior.
Finally, the nonlinear behavior of the frequency dependent Case II, representing a damreservoir
interaction system with a reflective boundary of finite extent, is studied in Figs. 6.16 and 6.17 which
present the effect of nonlinear behavior in the displacement and stiffness internal force, respectively.
Figure 6.16: Damreservoir interaction with finite extent reflective boundary (Case II).
Effect of nonlinearity in the displacement response.
Figure 6.17: Damreservoir interaction with finite extent reflective boundary (Case II).
Effect of nonlinearity in the stiffness force response.
In this case the nonlinear behavior is more important than in the radiation boundary case. As
explained before, the reflective boundary condition introduces energy in the system generating large
displacements even after the loading duration. Fig. 6.16 shows that all the peak values of the
displacement are considerable higher for the nonlinear behavior. It is important to notice that the
110
yielding branch of the stiffness force is reached only after time 5 s where the ground acceleration
loading is not present anymore.
One final observation valid for the three cases presented is that for this type of loading the nonlinear
behavior of the system influences only the magnitude but not the profile of the time history response.
Figs. 6.12, 6.14 and 6.16 prove that this is true no matter if the system is frequency dependent or
not.
111
112
For each type of 2D flow element a specific flow boundary element should be attached in the edges of
the fluid domain. A particular flux (or pressure derivative) in the direction perpendicular to the
reservoir edge is assigned to the flow boundary element, depending on the type of boundary
condition required to be modeled. As explained in Chapter 3, for the damreservoir interaction
problem these boundary conditions are bottom absorption, radiation boundary of infinite extent and
surface waves (See Section 3.1.3).
1
2
Straight and curved 2D flow boundary elements, B2HT and BC3HT, respectively, are shown in Fig.
7.2. The flux qy perpendicular to the boundary presents a linear variation along its length for B2HT
and a quadratic variation for BC3HT.
Similarly than for the twodimensional models, Fig. 7.3 shows two types of flow finite elements for
threedimensional models, namely HX8HT and CHX20H, respectively. BQ4HT and BCQ8HT are their
corresponding flow boundary elements which are shown in Fig. 7.4.
3
4
Two material parameters should be defined for the flow elements in a fluidstructure interaction
model. The first one is the conductivity (cF) of the fluid which in this type of analysis is a dummy
parameter which definition allows the calculation of the fluid domain conductivity matrix
. If in
addition the compressible effects are desired to be taken into account, then the sonic speed (c) should
be defined for the compressibility matrix
to be implemented (See Section 3.1.5).
One side of the fluidstructure interface elements are defined in terms of fluid pressure variables in a
direction perpendicular to the interaction surface, whereas the other side is defined in terms of
structure displacements variables in the three (or two) dimensions of analysis. For example, the
hybrid variables definition of the fluidstructure interface elements is shown in Fig. 7.6 for the line
interface BL4S2 and the plane interface BQ12S4.
One material parameter is required by DIANA for the definition of a fluidstructure interface property,
which is the fluid density ( ).
The definition of these parameters is used in the calculation of the radiation matrix
3.1.5).
(See Section
8
9
For detailed information about boundary surface elements see TNO DIANA (2011).
For detailed information about total strain crack models see TNO DIANA (2011).
117
Figure 7.7: DIANA predefined tension softening functions for total strain crack model.10
Figure 7.8: DIANA predefined compression functions for total strain crack model. 11
10
11
The types of finite elements used for each component of the model are indicated in Table 7.1. The
dam and foundation elements are the typical quadrilateral (Q8EPS) and triangular (T6EPS) plane
strain elements with linear interpolation. The characteristics of the reservoir flow and fluidstructure
interface elements are explained in Section 7.1.
The values of the linear and nonlinear material parameters used in this case study for the dam, the
foundation and the reservoir (fluid domain and boundary parameters) are listed in Tables 7.2, 7.3 and
7.4, respectively.
Table 7.1: Type of DIANA finite elements used for the case study modeling
Component
Type
Element
Dam
Plane strain 2D
Q8EPS
T6EPS
Foundation
Plane strain 2D
Q8EPS
T6EPS
Reservoir
General flow 2D
Q4HT
Reservoir boundaries
Flow boundary 2D
B2HT
Reservoirdam interface
Fluidstructure interface 1D
BL4S2
119
Value / Type
Units
Modulus of elasticity
YOUNG
2.7 x 1010
N / m2
Poison modulus
POISON
1.67 x 101
Density
DENSIT
2.4 x 103
kg / m3
Rayleigh damping
RAYLEI
8.19456 x 101
1.38396 x 103
TENCRV
LINEAR
Tensile strength
TENSTR
3.8 x 106
N / m2
EPSULT
5.0 x 104
Compression curve
COMCRV
CONSTA
Compressive strength
COMSTR
3.5 x 107
N / m2
Value / Type
Units
Modulus of elasticity
YOUNG
2.5 x 1010
N / m2
Poison modulus
POISON
2.0 x 101
Density
DENSIT
kg / m3
Value / Type
Units
Conductivity
CONDUC
2.5 x 1010
N / m2
Sonic speed
CSOUND
1.44 x 103
m/s
Density
DENSIT
1.0 x 103
kg / m3
ALPHAB
ALPHAB
5.0 x 101
Parameter
For transient and HFTD analysis the El Centro earthquake ground acceleration signal shown in Fig.
7.10 is defined as the loading input. On the other hand, for direct frequency analysis the
corresponding ground acceleration spectrum shown in Fig. 7.11 is taken into account.
7.3.2. Eigenanalysis
The eigenanalysis of the dam was performed following the two procedures presented and explained
in Section 3.1.9. The first procedure, known as dry dam eigenanalysis, does not take into account
the influence of the reservoir in the calculations, and therefore, it solves the free vibration equation of
motion of the structures given by Eq. (3.73) in terms of the mass ( ) and stiffness ( ) matrices.
(3.73)
121
On the other hand, the second procedure, known as wet dam eigenanalysis, does take into account
the influence of the fluid reservoir but assuming that it is incompressible and therefore, without
radiation but reflective boundary conditions. In this approach the reservoir introduces in the free
vibration equation of motion, given by Eq. (3.75), a fluid added mass term ( ) defined by Eq. (3.72)
(See Chapter 3).
(3.72)
(3.75)
The dry dam eigenfrequencies and effective mass participation in both directions of analysis are
presented in Fig. 7.12 for the first 15 modes of vibration. The fundamental vibration frequency for the
dam without reservoir interaction is 3.33 Hz. Besides, it can be noticed that the first modes of
vibration concentrate more than 90% of the vibration mass in both directions, being the first and third
modes the most important for X and Y direction, respectively.
Figure 7.12: Eigenanalysis results for the empty reservoir or dry dam case
The total mass of the dam is equal to 9.6 x 106 kg. When the wet dam eigenanalysis is performed,
the calculated fluid added mass is equal to 5.42 x 10 6, which means a mass increment of more than
122
50%. This considerable additional mass makes the system more flexible and therefore the eigenfrequencies obtained are smaller than in the dry dam case.
Indeed, this is confirmed by the fundamental eigenfrequency for the wet dam case shown in Fig.
7.13, which is reduced to 2.62 Hz. However, even since the first mode, the effective mass
participating in the Xdirection is higher than 115%, which means that the fluid added mass actively
contributes in the dams horizontal vibration. This is not the case for the vertical vibration (Ydirection)
where the fluid mass has no participation at all, due to the vertical orientation of the damreservoir
interface.
Figure 7.13: Eigenanalysis results for the full reservoir or wet dam case
The remarkable influence of the fluid added mass in the eigenfrequency values is not replicated in
the mode shapes. Fig. 7.14 shows the modes shapes of the 1st, 2nd and 4th vibration modes in the Xdirection. It can be noticed that the first mode shape is not influenced by the added fluid mass, no
matter that the eigenfrequency experiences a considerable reduction. Small magnitude differences
can be appreciated in the second and fourth mode shapes.
It is due to these negligible differences that the HFTD method uses the mode shapes of the empty
reservoir eigenanalysis for the modesuperposition method.
123
Figure 7.14: First three mode shapes in the Xdirection for empty (eigendry) and full
(eigenwet) reservoir cases
Finally, the influence of the contribution of the initial conditions in the geometric stiffness matrix, and
therefore in the eigenanalysis results, was investigated. For this purpose, the gravity load of the dam
and the hydrostatic pressure applied in the dams face and in the soil below the reservoir were
included as initial conditions.
124
As shown in Figs. 7.15 and 7.16 for the dry and wet dam cases, respectively, the influence of the
initial loading conditions in the stiffness matrix of the structure is not significant. For this reason, the
eigenanalysis results presented in Figs. 7.15 and 7.16 are almost the same as those presented in
Figs. 7.12 and 7.13.
Figure 7.15: Eigenanalysis results for the empty reservoir or dry dam case including
initial loading conditions
125
Figure 7.16: Eigenanalysis results for the full reservoir or wet dam case including initial
loading conditions
Case
Case
Case
Case
Case
I: Incompressible fluid.
II: Compressible fluid with reflective reservoir boundaries.
III: Compressible fluid with bottom absorption boundary.
IV: Compressible fluid with infinite extent radiation boundary.
V: Compressible fluid with bottom absorption and infinite extent radiation boundaries.
The ground acceleration spectrum shown in Fig. 7.11 is applied to the five systems shown Fig. 7.17,
and as a consequence, the AmplitudeFrequency response plots given by Figs. 7.18 to 7.21 give an
12
illustrative insight of the effect that compressibility and reservoir boundary conditions have in the
dynamic behavior of damreservoir interaction systems.
Figs. 18 and 19 show the relative Displacement AmplitudeFrequency plot, in the X and Y directions
respectively, of the dams crest node (higher point of the dam).
0.014
C1_incomp
0.012
C2_compres
C3_compres_botabs
0.01
C4_compres_infext
0.008
C5_compres_botabs_infext
0.006
0.004
0.002
0
0
10
11
Frequency (1/s)
Figure 7.18: Xdirection Displacement AmplitudeFrequency plot of the dams crest.
127
0.0045
C1_incomp
0.004
C2_compres
0.0035
C3_compres_botabs
0.003
C4_compres_infext
0.0025
C5_compres_botabs_infext
0.002
0.0015
0.001
0.0005
0
0
10
11
Frequency (1/s)
Figure 7.19: Ydirection Displacement AmplitudeFrequency plot of the dams crest.
For all the case studies the higher amplitudes are obtained around the fundamental vibration
frequency of the wetdam system equal to 2.62 Hz (See Figs. 7.13 and 7.16).
Comparing the cases C1 and C2 it can be concluded that assuming that the fluid of the reservoir is
compressible produces an increment of the displacement amplitudes. The opposite occurs in case C3
for which a bottom absorption boundary condition is introduced. This is due to the additional damping
introduced to the system by the bottom absorption (See Chapter 3).
On the other hand, in this case study the introduction of infinite extent radiation boundary condition
(cases C4 and C5) does not influences decisively the magnitude of the responses around the
fundamental frequency of vibration. Only for the higher vibration frequencies around 6 Hz the
introduction of the infinite extent radiation boundary damps out the peaks of Cases C1 and C2. The
main reason for the negligible contribution of the infinite extent boundary in this case study is the
long extension of the reservoir, defined in the model and equal to 390 m.
Finally, Figs. 20 and 21 present the relative velocity and acceleration, respectively, of the dams crest
node in the Xdirection. The same observations made for the displacement response are applicable to
both velocity and acceleration amplitudes, with the only difference that the importance of
compressibility and infinite extent radiation boundary is considerably higher for acceleration (Fig. 21)
than for any other response.
128
0.25
C1_incomp
C2_compres
0.2
C3_compres_botabs
C4_compres_infext
0.15
C5_compres_botabs_infext
0.1
0.05
0
0
10
11
Frequency (1/s)
Figure 7.20: Xdirection Velocity AmplitudeFrequency plot of the dams crest.
3.5
C1_incomp
C2_compres
C3_compres_botabs
2.5
C4_compres_infext
2
C5_compres_botabs_infext
1.5
1
0.5
0
0
10
11
Frequency (1/s)
Figure 7.21: Xdirection Acceleration AmplitudeFrequency plot of the dams crest.
The ground acceleration loading shown in Fig. 7.10 is applied to the three case studies and their
responses are determined using the Newmark (average acceleration) and HFTD methods.
The type of cracking nonlinearity, considered exclusively for the dams concrete, is described in
Section 7.1.5, whereas the values of the concrete nonlinear parameters are presented in Table 7.2.
7.3.4.1.
A comparison of the structural concrete cracking nonlinear responses obtained in the dams crest
employing both Newmark and HFTD methods is presented in Figs. 7.23 to 7.25 for the relative
displacement, velocity and acceleration responses in the Xdirection, respectively.
The empty reservoir is an ideal case in which the influence of the dynamic interaction of the reservoir
with the dam is neglected. For this reason, the magnitudes of the responses obtained are small and
the system barely reaches the nonlinear behavior. Actually, most of the behavior shown in Figs. 7.23
to 7.25 corresponds to a linear elastic behavior.
The HFTD and Newmark results coincide very well until the time 2.5 s. After this time step, the slight
cracking of concrete begins and the difference between both responses starts to be noticeable. The
reason for this difference other is that, in spite of being small in both methods, the extension and
magnitude of concrete cracking in Newmark is greater than in HFTD.
Fig. 7.26 shows the cracking pattern of HFTD and Newmark methods for the last time step of the
simulation equal to 4.49 s. It can be noticed that as expected the small cracking zone is located for
both methods in the foot of the dam in the side of the reservoir. However, Newmark solution develops
a more extended cracking zone for which the cracking strains are also higher in comparison with the
130
HFTD solution. An explanation for the inherent different nonlinear performance of both methods is
developed in the full reservoir with incompressible fluid case (Section 7.3.4.2).
0.05
Newmark_nlcrack
0.04
HFTD_nlcrack
0.03
0.02
0.01
0
0.01
0.02
0.03
0.04
0
0.5
1.5
2.5
3.5
4.5
Time (s)
Figure 7.23: Empty reservoir concrete cracking nonlinear response. Dams crest
displacement Xdirection.
0.8
Newmark_nlcrack
0.6
HFTD_nlcrack
0.4
0.2
0
0.2
0.4
0.6
0.8
1
1.2
0
0.5
1.5
2.5
3.5
4.5
Time (s)
Figure 7.24: Empty reservoir concrete cracking nonlinear response. Dams crest velocity
Xdirection.
131
30
Newmark_nlcrack
20
HFTD_nlcrack
10
0
10
20
30
0
0.5
1.5
2.5
3.5
4.5
Time (s)
Figure 7.25: Empty reservoir concrete cracking nonlinear response. Dams crest
acceleration Xdirection.
Figure 7.26: Empty reservoir concrete cracking nonlinear response. Cracking pattern at
time 4.49 s.
7.3.4.2.
When the reservoir interaction assuming incompressible fluid is considered, the responses of the dam
become greater13. First a comparison of the structural linear elastic responses obtained in the dams
13
The influence of the reservoir interaction in the response of the dam was analytically presented for the onedimensional models of Chapter 6.
132
crest employing both Newmark and HFTD methods is presented in Figs. 7.27 to 7.29 for the relative
displacement, velocity and acceleration in the Xdirection, respectively.
0.1
Newmark_linelast
0.08
HFTD_linelast
0.06
0.04
0.02
0
0.02
0.04
0.06
0.08
0
0.5
1.5
2.5
3.5
4.5
5.5
Time (s)
Figure 7.27: Incompressible fluid linear elastic response. Dams crest displacement Xdirection.
1.5
Newmark_linelast
1
HFTD_linelast
0.5
0
0.5
1
1.5
0
0.5
1.5
2.5
3.5
4.5
5.5
Time (s)
Figure 7.28: Incompressible fluid linear elastic response. Dams crest velocity Xdirection.
133
40
Newmark_linelast
30
HFTD_linelast
20
10
0
10
20
30
40
0
0.5
1.5
2.5
3.5
4.5
5.5
Time (s)
Figure 7.29: Incompressible fluid linear elastic response. Dams crest acceleration Xdirection.
Figs. 7.27 to 7.29 show a very good agreement in the linear elastic solutions obtained with both
methods. However, small differences are appreciated in the peak values since time 3.5 s thereafter,
especially for the acceleration response.
As it is explained in the following lines for the nonlinear response, the reason for these differences is
related with the choice of implementing the HFTD method in combination with the mode superposition
method. Not all the modes of vibration but only fifteen are included in the HFTD analysis, and
therefore part of the solution contributing with the peaks response is excluded.
Furthermore, the loading input signal shown in Fig. 7.10 has a maximum sampling frequency equal to
25 Hz. This means that any excitation frequency contained in the loading, or any modal vibration
frequency of the system, which is greater to 25 Hz is not included in the HFTD response.
Before showing the results obtained for the concrete cracking nonlinear behavior of the dam, the
hydrodynamic pressures generated by the linear elastic damreservoir interaction is shown for
different time steps in Fig. 7.30. It is clear that the hydrodynamic pressures can assume several types
of height distribution depending of the acceleration in the interface nodes for each time step of the
analysis.
134
Now, a comparison of the structural concrete cracking nonlinear responses obtained in the dams
crest, employing both Newmark and HFTD methods, is presented in Figs. 7.31 to 7.33 for the relative
displacement, velocity and acceleration in the Xdirection, respectively.
As mentioned before, the consideration of a full reservoir with incompressible fluid in the seismic
analysis produces a significant increment of the responses amplitudes. For example, Fig. 7.31 shows
that the maximum displacement obtained in this case study is close to between 0.06 m and 0.08 m.
On the other hand, Fig. 7.23 shows that when the reservoir interaction with the dam is neglected, the
maximum nonlinear displacement is less than 0.04 m. This means that the level of cracking is also
intensified when the incompressible fluid is taken into account, and as a consequence the difference
between the cracking nonlinear behaviors predicted by HFTD and Newmark methods is more obvious
than in the empty reservoir case.
135
0.1
Newmark_nlcrack
0.08
HFTD_nlcrack
0.06
0.04
0.02
0
0.02
0.04
0.06
0.08
0
0.5
1.5
2.5
3.5
4.5
5.5
Time (s)
Figure 7.31: Incompressible fluid concrete cracking nonlinear response. Dams crest
displacement Xdirection.
1.5
Newmark_nlcrack
1
HFTD_nlcrack
0.5
0
0.5
1
1.5
0
0.5
1.5
2.5
3.5
4.5
5.5
Time (s)
Figure 7.32: Incompressible fluid concrete cracking nonlinear response. Dams crest
velocity Xdirection.
136
40
Newmark_nlcrack
30
HFTD_nlcrack
20
10
0
10
20
30
40
0
0.5
1.5
2.5
3.5
4.5
5.5
Time (s)
Figure 7.33: Incompressible fluid concrete cracking nonlinear response. Dams crest
acceleration Xdirection.
Figs. 7.31 to 7.33 indicate that the nonlinear behavior described by the HFTD method is moderate,
resembling what is a known as a mild type of nonlinearity. For this reason, the shape of the timehistory plots is similar to the ones of the linear elastic behavior (Figs. 7.27 to 7.29), but with slightly
smaller amplitudes.
On the other hand, the nonlinear behavior described by the Newmark method is more severe,
diminishing considerably the response amplitudes and altering the periods of oscillations. The
substantial difference between the nonlinear cracking behaviors predicted by each method at
different time steps is illustrated in Fig. 7.34.
It can be noticed that in the time step 1.42 s, where the first crack appears, there is a good
agreement between both methods about which is the zone with higher tensile stresses. However,
during the following time steps, and especially around 2.5 s, the Newmark method develops a very
pronounced crack in the base due to the concentration of high tensile strains nearly that zone.
In spite of calculating a similar displacement field as Newmark does, HFTD method does not predict
such a tensile strain concentration in the base of the dam. Therefore, the HFTD tensile strain field is
uniformly distributed through the entire dams body and has considerably smaller magnitudes.
This observation about the displacement and strain fields obtained with both methods is graphically
demonstrated in Fig. 7.35 for the time step 2.40 s, when the big cracks shown in Fig. 7.34 for the
time step 2.61 s have not appeared yet.
137
Figure 7.34: Incompressible fluid reservoir nonlinear response. Dams concrete cracking
pattern at different time steps.
The large crack in the base of the dam calculated by the Newmark method (Fig. 7.34) is the reason
why the responses shown in Figs. 7.31 to 7.33 are highly nonlinear for this method. What is more,
due to the position and magnitude of the concrete cracking, the dam acts like an isolated system in
which the ground acceleration loading is not fully transmitted to the structure. For this reason, since
the time step 2.5 s the magnitudes of the responses in the crest of the dam calculated with the
Newmark method diminish substantially.
On the other hand, due to the fact that in the HFTD method the nonlinear response of the dam is
calculated based on the superposition of the linear elastic modes of vibration, the higher tensile
strains are predicted to be located in the vertical face of the dam interacting with the reservoir and
not in the base. In this case, the cracking is more distributed and therefore less severe. Furthermore,
138
the location of the higher tensile zones seems to be ruled by the first and the third mode shapes of
vibration in the horizontal direction (See Fig. 7.14).
7.3.4.3.
The frequency dependent damreservoir interaction system with compressible fluid and infinite extent
radiation boundary condition (Case III in Fig. 7.22) is directly solved in DIANA only by the Direct
Method (linear frequency domain solution) and the HFTD method (nonlinear transient solution). Due
to the frequency dependence of the system, Newmark method is not applicable for this case.
Therefore, nonlinear concrete cracking HFTD solution is presented in Figs. 7.36 to 7.38 for the dams
crest relative displacement, velocity and acceleration, respectively, in the Xdirection. A general
observation of the time history graphics allows concluding that for this particular model the inclusion
of compressibility in combination with the radiation boundary of infinite extent does not change
substantially the peak values of the nonlinear responses of the system. This analyzed in detail in the
next section.
139
0.1
HFTD_nlcrack
0.08
0.06
0.04
0.02
0
0.02
0.04
0.06
0.08
0
0.5
1.5
2.5
3.5
4.5
Time (s)
Figure 7.36: Compressible fluid with radiation boundary. HFTD nonlinear solution for the
dams crest displacement in Xdirection.
1.5
HFTD_nlcrack
1
0.5
0
0.5
1
1.5
0
0.5
1.5
2.5
3.5
4.5
Time (s)
Figure 7.37: Compressible fluid with radiation boundary. HFTD nonlinear solution for the
dams crest velocity in Xdirection.
140
30
HFTD_nlcrack
20
10
0
10
20
30
0
0.5
1.5
2.5
3.5
4.5
Time (s)
Figure 7.38: Compressible fluid with radiation boundary. HFTD nonlinear solution for the
dams crest acceleration in Xdirection.
7.3.4.4.
The influence of damreservoir interaction and fluid compressibility in the dams concrete cracking
nonlinear response is analyzed based on the HFTD solutions of the three cases previously studied
(Fig. 7.22). The dams crest relative displacement, velocity and acceleration in the Xdirection are
presented in Figs. 7.39 to 7.41, respectively.
0.1
C1_Empty reservoir
0.08
C2_Incompressible
0.06
C3_Compressible
0.04
0.02
0
0.02
0.04
0.06
0.08
0
0.5
1.5
2.5
3.5
4.5
Time (s)
Figure 7.39: Influence of the damreservoir interaction and fluid compressibility in the
concrete cracking nonlinear response. Dams crest displacement Xdirection.
141
1.5
C1_Empty reservoir
C2_Incompressible
C3_Compressible
1
0.5
0
0.5
1
1.5
0
0.5
1.5
2.5
3.5
4.5
Time (s)
Figure 7.40: Influence of the damreservoir interaction and fluid compressibility in the
concrete cracking nonlinear response. Dams crest velocity Xdirection.
40
C1_Empty reservoir
C2_Incompressible
C3_Compressible
30
20
10
0
10
20
30
40
0
0.5
1.5
2.5
3.5
4.5
Time (s)
Figure 7.41: Influence of the damreservoir interaction and fluid compressibility in the
concrete cracking nonlinear response. Dams crest acceleration Xdirection.
The consideration of the reservoir interaction in the seismic analysis is demonstrated to be important.
Fig. 7.39 shows that the maximum displacement of the full reservoir analysis (Cases C2 and C3) is
twice the maximum displacement of the empty reservoir analysis (case C1). Additionally, the vibration
period of the full reservoir cases is longer than the oscillation period of the empty reservoir case.
142
On the other hand, for this particular case compressibility does not play a fundamental role. Even
though higher displacement amplitudes are indeed obtained for the compressible fluid case, the
difference with the amplitudes of the incompressible fluid case is very small. Fig. 7.39 also shows that
the oscillation period of the compressible fluid case becomes progressively longer than the oscillation
period of the incompressible fluid case.
Anyhow, the time history acceleration plot shown in Fig. 7.41 demonstrates that the consideration of
the reservoir interaction and the fluid compressibility in the seismic analysis of dams produces very
different nonlinear dynamic responses. This is confirmed by the cracking pattern at different time
steps shown in Fig.7.42 for each of the three case studies.
Figure 7.42: Influence of the damreservoir interaction and fluid compressibility in the
nonlinear response. Dams concrete cracking pattern at different time steps.
Fig. 7.42 shows that when the reservoir interaction is not taken into account, the dam barely cracks;
there is only a very small cracking zone located in the dam foot. On the other hand, the first crack
appears at time step 1.42 s for the full reservoir with incompressible fluid case study. This is the case
that suffers the higher levels of cracking in the zone of the vertical damreservoir interface. Finally,
the higher levels of cracking in the inclined back face of the dam correspond to the compressible fluid
case study.
143
144
8.
8.1. Conclusions
Through the different parts of this thesis it has been demonstrated that the HFTD method is founded
on a strong theoretical basis which fundamentals is the frequency domain analysis, particularly, the
Fourier analysis.
The results of the SDOF numerical examples performed in Chapter 5 validate the high accuracy of the
HFTD method for linear and even for highly nonlinear systems, including simultaneous hysteretic
stiffness and damping nonlinearity, subject to several types of loading.
For linear elastic SDOF systems, the HFTD method results are almost exact for all types of loading,
with exception of sudden loading functions which starts at an initial value different than zero (f.e.:
cosine loading function). In these latter cases the results obtained are very accurate but not exact.
For nonlinear SDOF systems, if the size of the time step is appropriate, the HFTD method also shows
a great precision, similar to the accuracy of the Newmark method. In order to achieve very accurate
results, both methods require that the time step size is in the order of the natural period of vibration
of the system divided over one hundred.
However, in addition to the size of the time step, the research work has shown that the accuracy of
the HFTD method depends on two numerical implementation issues: 1) Time segmentation approach;
and 2) Extended Fourier Transform period.
The division of the response time span of interest in a determined number of segments is a procedure
implemented in the HFTD method to achieve convergence in nonlinear problems. This
implementation issue, denominated time segmentation approach, has proved to be very useful
especially when the system is highly nonlinear, the time span of interest is long, or the system
includes frequency dependent properties. Besides, the division of the time span of interest in a higher
number of time segments improves the accuracy of the results pertaining to the initial time steps.
The extension of the Fourier Transform period by adding a band of zeros, called quiet zone, to the
loading signal is a fundamental implementation issue of the HFTD method, which makes possible to
satisfy the initial conditions and determine the transient response of the system, instead of the
steadystate response. The properly definition of the HFTD method must perform all the Fourier
Transforms using this extended period which resembles the infinite period of the analytical Fourier
Integral.
The HFTD method for MDOF systems was implemented in DIANA in combination with the Mode
Superposition method. This implementation option, explained in detail in Chapter 4, is theoretically
consistent due to the fact that the HFTD method solves a pseudolinear equation in the frequency
domain.
The programming of the HFTD method for SDOF systems in MATLAB was essential for the
understanding, implementation and validation of the HFTD implementation for general applications in
DIANA.
145
8.
The application of the HFTD method is mainly interesting in MDOF systems with frequency dependent
properties, like the damreservoir interaction system subject to a seismic ground acceleration loading.
In Chapter 7 a case study of this system was tested in DIANA using the HFTD and Newmark methods.
For the linear elastic analysis of the damreservoir interaction system, the results obtained with the
HFTD and the Newmark methods show a very good agreement, in terms of accuracy and
performance. This is not the case for the concrete cracking nonlinear behavior of the dam, which
HFTD and Newmark results show considerable differences.
The main reason for the differences in the results when dealing with the concrete cracking nonlinearity of this particular case study is the great influence that the Mode Superposition method has in
the description of the nonlinear behavior of the structure determined with the HFTD method. Due to
the combination these two methods, the HFTD nonlinear response of the structure is calculated
based on the superposition of the linear elastic modes of vibration, and therefore, the distribution of
higher stresses and cracking zones is ruled by their associated mode shapes.
A totally different situation is defined by the Newmark method, which solution is not based on the
Mode Superposition method, but on the solution of the complete system. Logically, the consequences
of this inherent difference between both methods cannot be evaluated with the SDOF systems studied
in Chapter 5, which total response is exactly the same as the modal response.
Therefore, one important conclusion of the research work is that the best performance of the HFTD
method, in combination with the Mode Superposition method, is obtained when it is applied to MDOF
systems with frequency dependent properties and linear or mild nonlinear behavior.
Mild nonlinear behavior is understood as a type of slight nonlinearity which is not concentrated in
particular zones but spread in the whole structure. It is assumed that mild nonlinearity does not
modify drastically the structural configuration of the linear elastic system, and consequently, the
superposition of the modal responses system can still provide a good estimation of the nonlinear
response.
In the particular application of HFTD in damreservoir interaction systems, the frequency dependent
properties are composed by the fluid compressibility and the radiation boundary conditions of the
reservoir, like for example, the bottom absorption and the infinite extent.
From the thesis it can be concluded that these frequency dependent properties actually have an
impact in the response of the damreservoir interaction system. However, it was not possible to
outline any definitive conclusion about the relative importance of their inclusion in the systems
response.
Nevertheless, it is clear that the radiation boundary conditions, especially the bottom absorption, act
like additional damping in the system, and therefore, they reduce the magnitude of the systems
response.
The effect of fluid compressibility, on the other hand, is less predictable. For linear elastic frequency
domain analysis, the assumption of a compressible fluid clearly increases the amplitudes of the
response. However, for this particular case study, the time domain response obtained with the HFTD
method shows that, even though fluid compressibility indeed produces higher peaks, this increment is
small and constitutes a minor change in the timehistory response of the system.
One important advantage of the HFTD method is its numerical stability. Accurate solutions are
obtained without convergence problems mainly because the solution of the pseudolinear equation in
146
8.
the frequency domain does not require the calculation of a tangent stiffness matrix for each iteration
or each time step, as it is the case of the NewtonRaphson method. On the contrary, the HFTD
method uses the same linear elastic matrices through all the process, changing only the values of the
excitation frequencies, the excitation forces and the pseudo force vector. This allows the HFTD
method to solve large 3D models achieving convergence in an efficient way.
As a final conclusion, the HFTD method shows promising perspectives for its application in the
academic and professional practice, especially for the seismic analysis of fluidstructure and soilstructure interaction systems, or any other area for which the research of dynamic nonlinear behavior
of frequency dependent systems becomes every time more important.
8.2. Recommendations
It is recommended that the current implementation of the HFTD method in DIANA be used in the
investigation of different types of frequency dependent systems.
Threedimensional models of damreservoir interaction systems could be elaborated not only to keep
evaluating the performance of the HFTD method, but to study more deeply the importance of fluid
compressibility and radiation boundary conditions in the time response of this kind of systems.
It would be important to test the HFTD method using different types of nonlinear behavior (concrete
crushing, steel yielding, base sliding), and evaluate the results obtained in the context of the mildnonlinearity concept and the influence that the Mode Superposition method has on the HFTD
method.
The effect that different types of structural geometries have in the HFTD nonlinear responses
constitutes an interest research field which has not been studied yet. For some types of structural
geometries, like for example slender structures, the principal mode shapes describe more precisely the
actual deformation and stress distribution of the system, even after the incursion in the nonlinear
behavior stage.
Finally, for future research it is recommended to study the feasibility, in terms of computational
efficiency, of implementing the HFTD method without including the Mode Superposition method. This
would allow overcoming the restriction to mild nonlinear behavior, but assuming the cost of
increasing drastically the size of the system of equations to be solved.
147
8.
148
References
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
Aprile, A., Benedetti, A., and Trombetti, T. (1994). On nonlinear dynamic analysis in the
frequency domain: algorithms and applications . Earthquake Engineering and Structural
Dynamics, Vol.23, pp. 363388.
Clough, R.W., and Penzien, J. (2003). Structural dynamics. Computers and Structures,
California, Berkeley.
Craig, R.R., and Kurdila, A.J. (2006). Fundamentals of structural dynamics. John Wiley and
Sons, New Jersey.
Chavez, J.W., and Fenves, G.L. (1993). Earthquake analysis and response of concrete gravity
dams including base sliding. Report UBC/EERC93/07, Earthquake Eng. Research Center, Univ.
of California, Berkeley.
Chavez, J.W., and Fenves, G.L. (1994). EAGSLIDE: A computer program for the earthquake
analysis of concrete gravity damsincluding base sliding. Report UBC/SEMM94/02, Structural
Engineering Mechanics and Materials, Department of Civil Engineering, Univ. of California,
Berkeley.
Chopra, A.K. (2007). Dynamics of structures: theory and applications to earthquake
engineering. Third edition. Pearson Education Inc., New Jersey.
Darbre, G.R., and Wolf, J.P. (1988). Criterion of stability and implementation issues of hybrid
frequencytime domain procedure for nonlinear dynamic analysis. Earthquake Engineering and
Structural Dynamics, Vol.16, pp. 569581.
Darbre, G.R. (1990). Seismic analysis of a nonlinearly baseisolated soilstructure interacting
reactor building by way of the hybrid frequencytime domain procedure. Earthq. Eng. and
Struct. Dyn., Vol.19, pp. 725738.
Darbre, G.R. (1996). Nonlinear damreservoir interaction analysis. Proc. Eleven World Conf. on
Earthquake Engineering, Acapulco, Paper No. 760.
Fenves, G.L., and Chavez, J.W. (1990). Hybrid frequencytime domain analysis of nonlinear
fluidstructure systems. Proc. Fourth U.S. Nat. Conf. on Earthquake Engineering, California, pp.
97104.
Fenves, G.L., and Chopra, A.K. (1983). Effects of reservoir bottom absorption on earthquake
response of concrete gravity dams. Earthquake Engineering and Structural Dynamics, Vol.11,
pp. 809829.
Fenves, G.L., and Chopra, A.K. (1984). Earthquake analysis and response of concrete gravity
dams. Report UBC/EERC84/10, Earthquake Eng. Research Center, Univ. of California, Berkeley.
Humar, J.L. (2002). Dynamics of structures. Second edition. Balkema Publishers, Lisse.
Mansur, W.J. et al. (2000). Timesegmented frequencydomain analysis for nonlinear multidegreeoffreedom structural systems. Journal of Sound and Vibration, Vol. 237(3), 457475.
Metrikine, A.V., and Vrouwenvelder, A.C. (2006). Dynamics of structures Part 2: Wave
dynamics. Lecture Notes of Course CIE4140, Section of Structural Mechanics, Faculty of Civil
Engineering and Geosciences, Delft University of Technology, Delft.
Paz, M., and Leigh, W. (2004). Structural dynamics: theory and computation. Fifth edition.
Springer Science+Bussiness media, New York.
Rizos, D.C., and Karabalis, D.L. (2000). Fluidsoilstructure interaction. International series on
advances in earthquake engineering. WIT Press, UK.
The Mathworks, Inc. (2012). MATLAB and Statistics Toolbox Release 2012b, Natick,
Massachusetts.
TNO DIANA BV (2011). DIANA Users Manual Release 9.4.4, Delft.
149
20.
21.
22.
23.
References
Veletsos, A.S., and Ventura, C.E. (1985). Dynamic analysis of structures by the DFT method. J.
Struct. Eng. ASCE, Vol.111, 26252642.
Wilson, E.L. (2002). Threedimensional static and dynamic analysis of structures. Third Edition.
Computers and Structures Inc., California.
Zienkiewicz, O.C., and Bettes, P. (1978). Fluidstructure dynamic interaction and wave forces.
An introduction to numerical treatment. Int J Numer Meth Eng, Vol. 13, 116.
Zienkiewicz, O.C., Taylor, R.L., and Zhu, J.Z. (2005). The finite element method Its basis and
fundamentals. Sixth Edition. Elsevier ButterworthHeinemann, London.
150
START
HFRESP
FINISH
151
START
Fourier period
NLODTI = Total number of factors for the Fourier period
f1 = Free vibration fundamental frequency of the structure
= Damping ratio of the structure
TFOURI = max {TACSIG + 0.75/(f1 * ); 50 / f1}
IF (MOD (TFOURI, DT) .NE. 0 ) THEN
NLODTI = (TFOURI / DT) +1
ELSE
NLODTI = TFOURI / DT
END IF
TFOURI = NLODTI * DT
% Fourier period
152
Interaction matrix
FREQUENCY LOOP
]
]
% Real part
% Imaginary part
]
[
% Real part
% Imaginary part
153
FREQUENCY LOOP
YES
NEXT EXCITATION
FREQUENCY?
NO
Initialize response and pseudo force vectors
DISPLA = 0
VELOCI = 0
ACCELE = 0
PSEUFO = 0
GENPSF = 0
]
]
]
DOTISE
FINISH
154
START
ITIMSE = 1, NTIMSE
TDCAYF = NDCAYF * DT
[
[
% NDCAYF NSAMPL
]
]
]
TIMSEG
Output calculation
Dynamic fluid pressures
Strains and stresses of linear elements
YES
NEXT TIME
SEGMENT?
NO
FINISH
155
START
CONVTI = FPTISE  1
IMODE = 1, NMODES
Add decay function to the generalized pseudo force vector
MODE LOOP
PFDCYF =
Values of the decay function for generalized pseudo force (NMODES x NDCAYF)
[
]
CONVERGENCE LOOP
FOUGPF =
[
]
YES
NEXT MODE?
NO
FREQUENCY LOOP
Solution of pseudo linear system and determination of structural response in the frequency
domain
[
YES
NEXT EXCITATION
FREQUENCY?
NO
156
IMODE = 1, NMODES
MODE LOOP
YES
[
[
[
]
]
]
NEXT MODE?
NO
CONVERGENCE LOOP
DISPLA (ITIME)
VELOCI (ITIME)
ACCELE (ITIME)
Calculate error
GPSFER =
157
CONVERGENCE LOOP
GPSFER GPSTOL
&
CONVTI = ITIME  1
NO
YES
Update the converged time step counter
CONVTI = ITIME
YES
NEXT TIME
STEP?
NO
NO
CONVTI = LPTISE
YES
FINISH
158
159
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
HFTD for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
This script execute the HFTD method for linear and nonlinear simplfied
%
dynamic models of one or more degrees of freedom (DOF) subjected to
%
base (ground) acceleration loading.
%
%
After defining the system properties, the load signal and the HFTD
%
method parameters, the following results are obtained: time history
%
displacement, velocity, acceleration and nonlinear internal forces of
%
the system. Besides, the frequency response of the pseudo linear system
%
is also obtained.
%
%
The ground acceleration signal should be saved with the name "signal"
%
in the work folder as a matlab variable file called "signal.mat". The
%
values of the ground acceleration stored in the variable signal should
%
be normalized with respect to the gravity (g).
%
%
A Frequency dependent damping relation for the dashpots coefficients
%
can be specified in the function "f2_freq_dep_damp1".
%
%
The desired type of decaying function should be defined in the
%
function file "f5_decay_func".
%
%
The type of stiffness and damping nonlinearity to be used in the
%
problem should be selected in Section 16.6.6 of this script.
%
The selected function file is used to determine FNONLI.
%
The nonlinear parameters of each type of nonlinear function should be
%
defined inside the corresponding function file.
%
%
The output results are saved as excel tables, figures and matlab data
%
files. The file path where these output results files are saved should
%
be indicated Secttion 17 of this script and in the function files
%
"f15_excel_tables_v2" and "f16_plot_resp_v2".
%
%
160
disp('==========================================================================')
disp('
Problem name
')
disp('==========================================================================')
name=input('\nEnter the name of the file (this name is used to save the output data)... ','s');
disp('
')
pause(1);
161
KELAST=input('\nEnter the linear elastic stiffness coefficients k of the springs \n(in a row
vector enter NDOF+1 stiffness coefficients) .... ');
sizek=size(KELAST);
if sizek(1)~=1
disp(' ')
disp('Error message: The spring coefficients should be arranged in a row vector')
pause(2);
elseif sizek(2)~=NDOF+1
disp(' ')
disp('Error message: The row vectors defining the coefficientes of K should have NDOF+1
components ')
pause(2);
else
lock=1;
end
end
disp(' ')
disp('The linear elastic stiffness coefficients are defined to be')
KELAST
disp(' ')
pause(1);
disp('==========================================================================')
disp('
Definition of the damping matrices
')
disp('==========================================================================')
pause(1);
lock=0;
while lock==0
CELAST=input('\nEnter the constant (not frequency dependent) linear elastic damping
coefficients c \nof the dashpots (in a row vector enter NDOF+1 damping coefficients)... ');
sizec=size(CELAST);
if sizec(1)~=1
disp(' ')
disp('Error message: The dashpots coefficients should be arranged in a row vector')
pause(2);
elseif sizec(2)~=NDOF+1
disp(' ')
disp('Error message: The row vectors defining the constant dashpot coefficients should
have NDOF+1 components ')
pause(2);
else
lock=1;
end
end
disp(' ')
disp('The constant (not frequency dependent) linear elastic damping coefficients are defined to
be ')
CELAST
pause(1);
lock=0;
while lock==0
162
RAYLEI=input('\nEnter the coeficients for the definition of Rayleigh damping matrix aM+bK
\nin a row vector [a b]. If there is no Rayleigh damping enter [0 0]... ');
sizer=size(RAYLEI);
if sizer(1)~=1
disp(' ')
disp('Error message: The Rayleigh coefficients should be arranged in a row vector')
pause(2);
elseif sizer(2)~=2
disp(' ')
disp('Error message: The row vectors defining the Rayleigh coefficientes a and b should
have 2 components ')
pause(2);
else
lock=1;
end
end
a=RAYLEI(1);
b=RAYLEI(2);
disp(' ')
disp('The Rayleigh damping coefficients are defined to be ')
a
b
pause(1);
lock=0;
while lock==0
HYSTER=input('\nEnter the hysteretic damping ratio. If there is no hysteretic damping enter
0... ');
sizeh=size(HYSTER);
if sizeh(1)==1 && sizeh(2)==1
lock=1;
else
disp(' ')
disp('Error message: The input value must be scalar')
end
end
disp(' ')
disp('The hysteretic damping ratio is defined to be ')
HYSTER
disp(' ')
pause(1);
disp('==========================================================================')
disp('
Gravity constant
')
disp('==========================================================================')
pause(1);
lock=0;
while lock==0
g=input('\nEnter the value of the gravity constant g .... ');
sizeg=size(g);
if sizeg(1)~= 1  sizeg(2)~=1
disp(' ')
163
164
M=zeros(NDOF);
for IDOF=1:NDOF
M(IDOF,IDOF)=MCOF(IDOF);
end
disp(' ')
disp('The mass matrix M is')
M
pause(1);
%Stiffness matrix
K=f1_linelast_mat(NDOF,KELAST);
disp(' ')
disp('The linear elastic stiffness matrix K is')
K
pause(1);
%Damping matrix
Cnp=f1_linelast_mat(NDOF,CELAST);
disp(' ')
disp('The constant (not frequency dependent) linear elastic non proportional (dashpots) damping
matrix Cnp is')
Cnp
pause(1)
Cray=a*M+b*K;
disp(' ')
disp('The constant (not frequency dependent) linear elastic Rayleigh damping matrix Cray is')
Cray
pause(1)
C=Cnp+Cray;
disp(' ')
disp('The total constant (not frequency dependent) linear elastic damping matrix C is')
C
pause(1)
Ch=HYSTER*K;
disp(' ')
disp('The hysteretic linear elastic damping matrix Ch is')
Ch
pause(2);
disp(' ')
')
165
Mgen=transpose(Phi)*M*Phi;
w1=sqrt(Natfreq(1,1));
T1=2*pi/w1;
pause(1);
disp('
')
disp('The eigenvectors matrix is ')
disp('
')
Phi
pause(1);
disp('
')
disp('The eigenvalues (squared natural circular frequencies) matrix is')
disp('
')
Natfreq
pause(1);
disp('
')
disp('The fundamental period of the system is defined to be')
disp('
')
T1
pause(1);
disp('
')
disp('The fundamental circular frequency of the system is defined to be')
disp('
')
w1
pause(1);
166
Ch1
Cfd1
pause(1);
C1=C+Ch1+Cfd1;
disp(' ')
disp('The total damping matrix (constant + hysteretic + frequency dependent) ')
disp('for the fundamental mode of vibration is determined to be ')
C1
pause(1);
C1gen=transpose(Phi(:,1))*C1*Phi(:,1);
xi1=(1/(2*w1))*(C1gen/Mgen(1,1));
disp('
')
disp('The damping ratio of the fundamental mode of vibration of the system is defined to be')
disp('
')
xi1
disp('
')
pause(2);
167
168
opt=input('\nSelect the option for the definition of the HFTD time step size (DT). Enter "0"
for DT to be \nequal to the signal step size (SIGNDT) or enter "1" for DT to be a multiple of
SIGNDT .... ');
if opt==0
TIMES=1;
lock=1;
elseif opt==1
pause(2);
TIMES=input('\nEnter the number of times that SIGNDT should be multiplied to obtain
DT.... ');
if mod(TIMES,1)~=0  TIMES<=0
disp('
')
disp('Error message: The input value should be a positive integer')
else
lock=1;
end
else
pause(2);
disp('
')
disp('Error message: The input value must be 0 or 1')
end
end
DT=TIMES*SIGNDT;
prob=[TACSIG+0.75*T1/abs(xi1),50*T1,2*TACSIG];
TFOURI=max(prob);
NLODTI=ceil(TFOURI/DT);
TFOURI=DT*NLODTI;
pause(1);
disp('
')
disp('The number of time points (NLODTI) contained in the Fourier period is calculated to be ')
disp('
')
NLODTI
pause(1);
disp('
')
disp('The HFTD time step size (DT) is defined to be')
disp('
')
DT
pause(1);
disp('
')
disp('The Fourier period (TFOURI) is calculated to be')
disp('
')
TFOURI
disp('
')
pause(2)
169
%
until the Fourier period is covered.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
LODTIM=zeros(NLODTI,1);
LODFAC=zeros(NLODTI,1);
for k=1:NLODTI
LODTIM(k)=(k1)*DT;
j=1+(k1)*TIMES;
if j<=NFACT
LODFAC(k)=SIGFAC(j);
else
LODFAC(k)=0;
end
end
170
end
end
disp('
disp('The
be')
disp('
NFRESP
pause(1);
disp('
disp('The
disp('
TRESP
disp('
pause(2);
')
number of time steps defining the time span of interest length (NFRESP) is defined to
')
')
response time span of interest (TRESP) is defined to be')
')
')
171
disp('
')
NTIMSE=ceil(NFRESP/NSAMPL);
pause(1);
disp('The number of time segments (NTIMSE) through all the interest time span is defined to be')
disp('
')
NTIMSE
disp('
')
pause(2);
172
173
174
Section 14: Hysteretic and frequency dependent damping for each excitation frequency
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The hysteretic and frequency dependent damping matrices are
%
calculated for each excitattion frequency between HIPASS and
%
LOPASS. The frequency dependent damping should be defined by the
%
user in the function file "f2_freq_dep_damp1".
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
CFREDE=zeros(LOPASSHIPASS+1,NDOF+1);
Cfd=zeros(NDOF,NDOF,LOPASSHIPASS+1);
Chy=zeros(NDOF,NDOF,LOPASSHIPASS+1);
for IEXCIT=HIPASS:LOPASS
w=2*pi*EXCITA(IEXCIT);
[CFREDE(IEXCITHIPASS+1,:),EM]=f2_freq_dep_damp1(NDOF,w); %Function defining the frequency
dependent damping
Cfd(:,:,IEXCITHIPASS+1)=f1_linelast_mat(NDOF,CFREDE(IEXCITHIPASS+1,:));
if w==0
Chy(:,:,IEXCITHIPASS+1)=0*Ch;
else
Chy(:,:,IEXCITHIPASS+1)=(1/w)*Ch;
end
end
175
Section 16.2: Ground acceleration decay function definition for the current segment
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The function "f5_decay_func" is called in order to determine
%
the ground acceleration factors pertaining to the decay
%
function appended at the end of the current time segment.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
[GADCYF]=f5_decay_func(LODFAC,LPTISE,NDCAYF,DT,TDCAYF);
Section 16.3: Load factors for the Fourier transform of the ground acceleration
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The load factors used for the Fourier transform of the ground
%
acceleration vary in each time segment. The ground acceleration
%
load factors including the decaying function used for the
%
current time segment are stored in the variable SEGFAC.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
SEGFAC=LODFAC;
SEGFAC(LPTISE+1:LPTISE+NDCAYF)=GADCYF;
SEGFAC(LPTISE+NDCAYF+1:NLODTI)=zeros;
176
177
fprintf('Starting the iteration process untill convergence is achieved for all the time steps
\ninside segment number... %d\n',ITIMSE)
while convlock==0
iter=iter+1;
disp('
')
disp('Starting iteration number... ')
disp(iter)
disp('
')
disp('Fourier transform of the pseudo force vector ... ')
for IDOF=1:NDOF
Section 16.6.2: Load factors for the Fourier transform of the pseudo force
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The load factors used for the Fourier transform of the
%
pseudo force vary in each iteration in accordance with
%
the results obtained in the previous iteration. The
%
pseudo force load factors including the decaying
%
function used for the current time segment and
%
iteration are stored in the variable PSEUFO.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
PSEUFO(IDOF,LPTISE+1:LPTISE+NDCAYF)=PFDCYF(IDOF,:);
178
179
segment.
[IFOUAR(IDOF,:),IFOUAI(IDOF,:)]=f8_inv_fast_fourier_transf(ACCFRE(IDOF,:),NLODTI,FPTISE,LPTISE);
% Inverse Fourier transform of the acceleration vector for the time steps within the current
segment.
end
Section 16.6.6.1: Nonlinear or linear stiffness force, internal spring force and state variable
of time step 1
%[KEY(:,ITIME),SPRFOR(:,ITIME),FNLSTI(:,ITIME),UT(:,ITIME),UC(:,ITIME),UO(:,ITIME),EMKEY]=f9_lin_
stiff_int_force(NDOF,CEROS2,CEROS1,DISPLA(:,ITIME),CEROS2,KELAST);
%[KEY(:,ITIME),SPRFOR(:,ITIME),FNLSTI(:,ITIME),UT(:,ITIME),UC(:,ITIME),UO(:,ITIME),EMKEY]=f10_non
lin_hyster_stiff_int_force_v3(VELOCI(:,ITIME),NDOF,CEROS2,CEROS1,DISPLA(:,ITIME),CEROS2,CEROS2,CE
ROS2,CEROS2,ITIME,KELAST);
[KEY(:,ITIME),SPRFOR(:,ITIME),FNLSTI(:,ITIME),UT(:,ITIME),UC(:,ITIME),UO(:,ITIME),EMKEY]=f11_nonl
in_hyster_stiff_int_force_cubic(VELOCI(:,ITIME),NDOF,CEROS2,CEROS1,DISPLA(:,ITIME),CEROS2,CEROS2,
CEROS2,ITIME,KELAST);
Section 16.6.6.2: Nonlinear or linear damping force, internal dashpot force and state
variable of time step 1
180
%[FNLDAM(:,ITIME),DASHFO(:,ITIME)]=f12_lin_damp_int_force_v2(NDOF,VELOCI(:,ITIME),CEROS2,CELAST);
%Nonlinear damping force and internal dashpot force of time step i+1
[FNLDAM(:,ITIME),DASHFO(:,ITIME)]=f13_nonlinmet_damp_int_force(NDOF,DISPLA(:,ITIME),VELOCI(:,ITIM
E),CELAST);
else
Section 16.6.6.3: Nonlinear or linear stiffness force, internal spring force and state variable
of time step i+1
%[KEY(:,ITIME),SPRFOR(:,ITIME),FNLSTI(:,ITIME),UT(:,ITIME),UC(:,ITIME),UO(:,ITIME),EMKEY]=f9_lin_
stiff_int_force(NDOF,KEY(:,ITIME1),DISPLA(:,ITIME1),DISPLA(:,ITIME),SPRFOR(:,ITIME1),KELAST);
%[KEY(:,ITIME),SPRFOR(:,ITIME),FNLSTI(:,ITIME),UT(:,ITIME),UC(:,ITIME),UO(:,ITIME),EMKEY]=f10_non
lin_hyster_stiff_int_force_v3(VELOCI(:,ITIME),NDOF,KEY(:,ITIME1),DISPLA(:,ITIME1),DISPLA(:,ITIME),SPRFOR(:,ITIME1),UT(:,ITIME1),UC(:,ITIME1),UO(:,ITIME1),ITIME,KELAST);
[KEY(:,ITIME),SPRFOR(:,ITIME),FNLSTI(:,ITIME),UT(:,ITIME),UC(:,ITIME),UO(:,ITIME),EMKEY]=f11_nonl
in_hyster_stiff_int_force_cubic(VELOCI(:,ITIME),NDOF,KEY(:,ITIME1),DISPLA(:,ITIME1),DISPLA(:,ITIME),UT(:,ITIME1),UC(:,ITIME1),UO(:,ITIME1),ITIME,KELAST);
Section 16.6.6.4: Nonlinear or linear damping force, internal dashpot force and state
variable of time step i+1
%[FNLDAM(:,ITIME),DASHFO(:,ITIME)]=f12_lin_damp_int_force_v2(NDOF,VELOCI(:,ITIME)VELOCI(:,ITIME1),DASHFO(:,ITIME1),CELAST); %Nonlinear damping force and internal dashpot force of time step
i+1
[FNLDAM(:,ITIME),DASHFO(:,ITIME)]=f13_nonlinmet_damp_int_force(NDOF,DISPLA(:,ITIME),VELOCI(:,ITIM
E),CELAST);
end
if EMKEY~=100
return
end
Section 16.6.6.5: Total nonlinear force, linear force and pseudo force
FNONLI(:,ITIME)=FNLSTI(:,ITIME)+FNLDAM(:,ITIME); % Nonlinear internal restoring force
FLINEA(:,ITIME)=f14_linear_int_rest_force(K,DISPLA(:,ITIME),Cnp,VELOCI(:,ITIME)); %
Linear internal restoring force vector
PSEUFO(:,ITIME)=FLINEA(:,ITIME)FNONLI(:,ITIME); % Pseudo force vector
181
end
182
disp(NOTCON)
elseif ITIMSE~=NTIMSE
disp('Convergence achieved for all the time steps within the current segment. Next
segment will be evaluated ... ')
else
disp('Finish: Convergence achieved for all the time steps of the interest span ')
end
end
end
183
O(IDOF,:),t,UT(IDOF,:),UC(IDOF,:),UO(IDOF,:),EXCITA,DISFRE(IDOF,1:MEXCIT),VELFRE(IDOF,1:MEXCIT),A
CCFRE(IDOF,1:MEXCIT));
f16_plot_resp_hftd_v2(IDOF,name,DISPLA(IDOF,:),VELOCI(IDOF,:),ACCELE(IDOF,:),SPRFOR(IDOF,:),DASHF
O(IDOF,:),DISREL(IDOF,:),VELREL(IDOF,:),t,EXCITA,DISFRE(IDOF,1:MEXCIT),VELFRE(IDOF,1:MEXCIT),ACCF
RE(IDOF,1:MEXCIT),FOUACC(1:MEXCIT));
else
f15_excel_tables_hftd_v2(IDOF,name,nullrow1,nullrow1,nullrow1,nullrow1,nullrow1,nullrow1,nullrow1
,nullrow1,KEY(IDOF,:),SPRFOR(IDOF,:),DASHFO(IDOF,:),t,UT(IDOF,:),UC(IDOF,:),UO(IDOF,:),EXCITA,nul
lrow2,nullrow2,nullrow2);
f16_plot_resp_hftd_v2(IDOF,name,nullrow1,nullrow1,nullrow1,SPRFOR(IDOF,:),DASHFO(IDOF,:),DISREL(I
DOF,:),VELREL(IDOF,:),t,EXCITA,nullrow2,nullrow2,nullrow2,nullrow2);
end
end
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Change the name of variables to be saved
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp(' ')
disp('Changing the name of the selected variables to be saved...')
%
The new names of the variables are created and stored using the
%
function "f17_new_name_var"
[vk,vc,vm,vtol,vg,vphi,vnatfreq,vt1,vw1,vxi1,vsignal,vdt,vnsampl,vntimse,vmexcit,vdeltaf,vexcita,
vlopass,vhipass,vnresit,vdispla,vveloci,vaccele,vpseufo,vnlstiff,vnldamp,vfnonli,vflinea,vkey,vsp
rinfor,vdashfor,vdisfre,vvelfre,vaccfre,vt,vut,vuc,vdisrel,vvelrel,vfouacc]=f17_new_name_var_hftd
_v2(name);
%
The new names are assigned to all the variables to be saved.
eval([vk ' = K;']);
eval([vc ' = C;']);
eval([vm ' = M;']);
eval([vtol ' = tol;']);
eval([vg ' = g;']);
eval([vphi ' = Phi;']);
eval([vnatfreq ' = Natfreq;']);
eval([vt1 ' = T1;']);
eval([vw1 ' = w1;']);
eval([vxi1 ' = xi1;']);
eval([vsignal ' = signal;']);
eval([vdt ' = DT;']);
eval([vnsampl ' = NSAMPL;']);
eval([vntimse ' = NTIMSE;']);
eval([vmexcit ' = MEXCIT;']);
eval([vdeltaf ' = DELTAF;']);
eval([vexcita ' = EXCITA;']);
eval([vlopass ' = LOPASS;']);
eval([vhipass ' = HIPASS;']);
eval([vnresit ' = NRESIT;']);
eval([vdispla ' = DISPLA;']);
eval([vveloci ' = VELOCI;']);
184
185
disp(' ')
disp('End of the HFTD program... Check the results files saved in the specified folder')
186
function [DCYFAC]=f5_decay_func(INPFAC,LPTISE,NDCAYF,DT,TDCAYF)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
HFTD for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to determine the factors of the decaying function appended to
%
the ground acceleration and pseudo force.
%
%
inputs
%
%
INPFAC Vector containing the input (ground acceleration or pseudo
%
force)factors for each time step.
%
LPTISE Index indicating the last time point of the current
%
segment.
%
NDCAYF Number of time steps defining the length of the decaying
%
function.
%
DT
Time step size.
%
TDCAYF Decaying function length.
%
%
outputs
%
%
DCYFAC Decaying function factors for the input (ground
%
acceleration or pseudo force). Different options for
%
decaying function expressions can be used by uncomment the
%
chosen one.
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
f0=INPFAC(LPTISE);
DCYFAC=zeros(1,NDCAYF);
for k=1:NDCAYF
DCYFAC(k)=f0*(1k*DT/TDCAYF);
%DCYFAC(k)=f0*(1k*DT/TDCAYF+sin(2*pi*k*DT/TDCAYF)/(2*pi));
end
187
function [FOUREA,FOUIMA]=f6_fast_fourier_transf(x,N,MEXCIT)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
HFTD for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to apply the direct discrete Fourier transform to a time
%
dependent variable
%
%
inputs
%
%
x
Time dependent variable (signal) to which direct Fourier
%
tranform is applied.
%
N
Number of time points contained in x.
%
MEXCIT Maximum number of frequency samples to be analyzed.
%
%
%
outputs
%
%
FOUREA Real part of the direct Fourier transform of signal x.
%
FOUIMA Imaginary part of the direct Fourier transform of signal x.
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% FFT computation
y = fft(x,N);
rey=real(y);
imy=imag(y);
% Scale FFT output
reh=rey/N;
imh=imy/N;
% Copy only the frequency responses that are going to be used in the
% frequency domain analysis. The conjugates are not required to be copied
% since they do not provide additional information.
FOUREA(1,:)=reh(1:MEXCIT);
FOUIMA(1,:)=imh(1:MEXCIT);
188
function
[DIFREA,DIFRIM]=f7_solver_pseudo_lin_direct_freq_v2(K,C,Cfd,Chy,M,w,NDOF,FOUACR,FOUACI,FOUPFR,FOU
PFI)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
HFTD for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to solve a linear system in the frequency domain using the
%
direct method
%
%
[(w^2)*M + i*w*C + K] u(w) = M*I*ug(w) + q(w)
%
%
inputs
%
%
K
Linear elastic stiffness matrix of the system.
%
C
Linear elastic constant damping matrix of the system.
%
Cfd
Linear elastic frequency dependent damping matrix of the
%
system.
%
Chy
Linear elastic hysteretic damping matrix of the system.
%
M
Mass matrix of the system.
%
w
Excitation frequency.
%
NDOF
Number of degres of freedom.
%
FOUACR Fourier transform of the ground acceleration (real part)
%
corresponding to the excitation frequency w
%
FOUACI Fourier transform of the ground acceleration (imaginary part)
%
corresponding to the excitation frequency w
%
FOUPFR Fourier transform of the pseudo force vector (real part)
%
corresponding to the excitation frequency w
%
FOUPFI Fourier transform of the pseudo force vector (imaginary part)
%
corresponding to the excitation frequency w
%
%
outputs
%
%
DIFREA Displacement vector in the frequency domain (real part)
%
obtained from the solution of the linear system of
%
equations.
%
%
DIFRIM Displacement vector in the frequency domain (imaginary
%
part) obtained from the solution of the linear system of
%
equations.
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Components of the modified stiffness matrix in the frequency domain
A=K(w^2)*M;
B=w*(C+Cfd+Chy);
% Assembling the modified stiffness matrix
189
190
function [IFOURE,IFOUIM]=f8_inv_fast_fourier_transf(x,N,FPTISE,LPTISE)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
HFTD for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to apply the inverse discrete Fourier transform to a time
%
dependent variable
%
%
inputs
%
%
x
Frequency dependent complex variable (spectrum) to which inverse Fourier
%
tranform is applied.
%
N
Number of frequency points contained in x.
%
FPTISE First time point of the signal that is copied as output
%
LPTISE Last time point of the signal that is copied as output
%
%
outputs
%
%
IFOURE Real part of the inverse Fourier transform of spectrum x.
%
IFOUIM Imaginary part of the inverse Fourier transform of spectrum x.
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% IFFT computation
y = ifft(N*x,N,'symmetric');
rey=real(y);
imy=imag(y);
% Copy only the results of the time steps within the current time segment
IFOURE=rey(FPTISE:LPTISE);
IFOUIM=imy(FPTISE:LPTISE);
191
function FLINEA=f14_linear_int_rest_force(K,DISPLA,C,VELOCI)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
HFTD for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the internal linear restoring force
%
%
%
inputs
%
%
K
Stiffness matrix of the system
%
DISPLA Displacement vector
%
C
Damping matrix of the system
%
VELOCI Velocity vector
%
%
%
outputs
%
%
FLINEA Vector containing the linear restoring force
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
FLINEA=K*DISPLA+C*VELOCI;
192
193
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Newmark method for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
This script execute the Newmark method for linear or nonlinear simplfied
%
dynamic models of one or more degrees of freedom (DOF's) subjected to
%
base (ground) acceleration.
%
%
After defining the system properties and the load signal, the following
%
results are obtained: time history displacement, velocity, acceleration
%
and nonlinear internal forces of the system.
%
%
The name of the function indicating the type of nonlinearity to be used
%
in the problem is indicated in Section 11. The functions describing
%
type of tangential stiffnes and damping matrix should be selected.
%
Besides, the nonlinear stiffness and damping models are defined inside
%
the function "f18_Newt_Raph_modf_v6".
%
%
The output results are saved as excel tables, figures and matlab data
%
files. The file path where these output results files are saved should
%
be indicated Section 12 of this script and in the function files
%
"f19_excel_tables_newm" and "f20_plot_resp_newm".
%
%
194
%
and linear acceleration cases are the two default options.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
disp('==========================================================================')
disp('
Define the Newmark method to be executed
')
disp('==========================================================================')
lock=0;
while lock==0
pause(2);
opt=input('\nSelect a Newmark special case or the numerical values of the parameters gamma
and beta. \nEnter "AA" for Average Acceleration or enter "LA" for Linear Acceleration, \nor "OC"
for other case .... ','s');
h=strcmp(opt,{'AA','LA','OC'});
if h(1)==1
gama=0.5;
beta=0.25;
lock=1;
elseif h(2)==1
gama=0.5;
beta=1/6;
lock=1;
elseif h(3)==1
pause(1);
gama=input('\nEnter the value of gamma) .... ');
if gama<0  gama>1
disp('
')
disp('Error message: Not valid specified value for gamma')
else
pause(1);
beta=input('\nEnter the value of beta) .... ');
if beta<0  beta>1
disp('
')
disp('Error message: Not valid specified value for beta')
else
lock=1;
end
end
else
pause(2);
disp('
')
disp('Error message: Option not correctly chosen. Choose one of the three available
options')
end
end
195
%
system. The gravity constant and the tolerance and maximum number of
%
iterations for nonlinear analysis are also input by the user.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
disp('==========================================================================')
disp('
Define the elastic stiffness, elastic damping and mass matrices
')
disp('==========================================================================')
pause(1);
lock=0;
while lock==0
MCOF=input('\nEnter the masses of each degree of freedom m_i \n(in a row vector enter one
mass per DOF).... ');
sizem=size(MCOF);
if sizem(1)~=1
disp(' ')
disp('Error message: The masses should be arranged in a row vector')
pause(2);
else
lock=1;
end
end
NDOF=sizem(2);
pause(1);
disp(' ')
disp('The number of degrees of freedom (NDOF) of the system is defined to be ')
NDOF
pause(1)
disp(' ')
disp('The masses of each degree of freedom are defined to be ')
MCOF
pause(1);
lock=0;
while lock==0
KELAST=input('\nEnter the linear elastic stiffness coefficients k_i of the springs \n(in a
row vector enter NDOF+1 stiffness coefficients) .... ');
sizek=size(KELAST);
if sizek(1)~=1
disp('
')
disp('Error message: The spring coefficients should be arranged in a row vector')
pause(2);
elseif sizek(2)~=NDOF+1
disp(' ')
disp('Error message: The row vectors defining the coefficientes of K should have NDOF+1
components ')
pause(2);
else
lock=1;
end
end
disp(' ')
196
197
if sizet(1)~= 1
disp(' ')
disp('Error
pause(2);
elseif tol>0.01
disp(' ')
disp('Error
pause(2);
else
lock=1;
end
 sizet(2)~=1
message: tol must be a scalar value')
 tol<=0
message: tol must be positive and smaller than 10^2')
end
disp(' ')
disp('The tolerance for convergence check is defined to be ')
tol
pause(1);
lock=0;
while lock==0
maxite=input('\nEnter the maximum number of iterations for convergence check in nonlinear
analysis .... ');
sizemax=size(maxite);
if sizemax(1)~= 1  sizemax(2)~=1
disp(' ')
disp('Error message: maxite must be a scalar value')
pause(2);
elseif mod(maxite,1)~=0  maxite<=0
disp(' ')
disp('Error message: maxite must be positive and integer')
pause(2);
else
lock=1;
end
end
disp(' ')
disp('The maximum number of iterations for convergence check is defined to be ')
maxite
pause(2);
198
%Mass matrix
M=zeros(NDOF);
for IDOF=1:NDOF
M(IDOF,IDOF)=MCOF(IDOF);
end
l=ones(NDOF,1);
P=M*l;
disp(' ')
disp('The mass matrix M is')
M
pause(1);
%Stiffness matrix
K=f1_linelast_mat(NDOF,KELAST);
disp(' ')
disp('The linear elastic stiffness matrix K is')
K
pause(1);
%Damping matrix
C=f1_linelast_mat(NDOF,CELAST);
disp(' ')
disp('The linear elastic damping matrix C is')
C
pause(2);
')
199
disp('
')
disp('The eigenvalues (squared natural circular frequencies) matrix is')
disp('
')
Natfreq
pause(1);
disp('
')
disp('The fundamental period of the system is defined to be')
disp('
')
T1
pause(1);
disp('
')
disp('The fundamental circular frequency of the system is defined to be')
disp('
')
w1
pause(1);
disp('
')
disp('The damping ratio of the fundamental mode of vibration of the system is defined to be')
disp('
')
xi1
disp('
')
pause(2);
200
if lock==1
disp('Error message:
return
elseif lock==2
disp('Error message:
return
elseif lock==3
disp('Error message:
disp('
')
pause(2);
fprintf('The problem
return
end
end
SIGFAC=g*signal(:,2);
TACSIG=NFACT*SIGNDT;
pause(2);
disp('
')
disp('The number of time points contained in the ground acceleration signal is ')
disp('
')
NFACT
pause(1);
disp('
')
disp('The time step size of the ground acceleration signal is ')
disp('
')
SIGNDT
pause(1);
disp('
')
disp('The duration of the ground acceleration signal is ')
disp('
')
TACSIG
pause(1);
disp('
')
disp('The maximum ground acceleration magnitude contained in the signal is ')
disp('
')
disp(max(abs(SIGFAC)))
disp('
')
pause(2);
201
lock=0;
while lock==0
pause(2);
opt=input('\nSelect the option for the definition of the HFTD time step size (DT). Enter "0"
for DT to be equal to the signal step \nsize (SIGNDT) or enter "1" for DT to be a multiple of
SIGNDT .... ');
if opt==0
TIMES=1;
lock=1;
elseif opt==1
pause(2);
TIMES=input('\nEnter the number of times that SIGNDT should be multiplied to obtain (DT)
.... ');
if mod(TIMES,1)~=0  TIMES<=0
disp('
')
disp('Error message: The input value should be a positive integer')
else
lock=1;
end
else
pause(2);
disp('
')
disp('Error message: The input value must be 0 or 1')
end
end
DT=TIMES*SIGNDT;
pause(1);
disp('
')
disp('The Newmark time step size (DT) is defined to be')
disp('
')
DT
pause(2)
202
')
number of time steps defining the time span of interest length (NFRESP) is defined to
')
')
time span of interest for determining the response (TRESP) is defined to be')
')
')
203
KEYSTI=zeros(NDOF+1,NFRESP);
KEYDAM=zeros(NDOF+1,NFRESP);
UT=zeros(NDOF+1,NFRESP);
UC=zeros(NDOF+1,NFRESP);
UO=zeros(NDOF+1,NFRESP);
NITERA=zeros(1,NFRESP);
204
disp('
')
end
end
pause(2)
205
,VELREL(IDOF,:),t)
end
end
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Changing the name of variables to be saved
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
disp(' ')
disp('Changing the name of the selected variables to be saved...')
%
The new names of the variables are created and stored using the
%
function f21_new_name_var_newm
[vk,vc,vm,vtol,vg,vphi,vnatfreq,vt1,vw1,vxi1,vsignal,vdt,vdispla,vveloci,vaccele,vnlstiff,vnldamp
,vkeystiff,vkeydamp,vsprinfor,vdashfor,vt,vnitera,vut,vuc,vdisrel,vvelrel]=new_name_var_newm(name
);
%
The new names are assigned to all the variables to be saved.
eval([vk ' = K;']);
eval([vc ' = C;']);
eval([vm ' = M;']);
eval([vtol ' = tol;']);
eval([vg ' = g;']);
eval([vphi ' = Phi;']);
eval([vnatfreq ' = Natfreq;']);
eval([vt1 ' = T1;']);
eval([vw1 ' = w1;']);
eval([vxi1 ' = xi1;']);
eval([vsignal ' = signal;']);
eval([vdt ' = DT;']);
eval([vdispla ' = DISPLA;']);
eval([vveloci ' = VELOCI;']);
eval([vaccele ' = ACCELE;']);
eval([vnlstiff ' = FNLSTI;']);
eval([vnldamp ' = FNLDAM;']);
eval([vkeystiff ' = KEYSTI;']);
eval([vkeydamp ' = KEYDAM;']);
eval([vsprinfor ' = SPRFOR;']);
eval([vdashfor ' = DASHFO;']);
eval([vt ' = t;']);
eval([vnitera ' = NITERA;']);
eval([vut ' = UT;']);
eval([vuc ' = UC;']);
eval([vdisrel ' = DISREL;']);
eval([vvelrel ' = VELREL;']);
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
The selected variables are saved
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
pause(1)
disp(' ')
disp('Saving selected set of variables...')
%File path where the data is store. The file path can be changed to another
206
%desired location.
path='C:\Users\slfx\Documents\Fernando\Graduation project\Master thesis\Analysis files\Matlab
models\Newmark_nonli_MDOF\Newm Output results\';
%path='C:\Users\Fernando Sirumbal\Documents\Documents FERNANDO\TU Delft\Academic\MSc
thesis\Master thesis\Analysis files\Matlab models\Newmark_nonli_MDOF\Newm Output results\';
%Name of the matlab data files
filename1=[path 'results_' name]; %Name of the matlab file containing the results variables
filename2=[path 'data_' name]; %Name of the matlab file conatining the data variables
%Saving the variables in both files
save(filename1,vdispla,vveloci,vaccele,vkeystiff,vkeydamp,vsprinfor,vdashfor,vt,vnlstiff,vnldamp,
vnitera,vut,vuc,vdisrel,vvelrel);
save(filename2,vk,vc,vm,vtol,vg,vphi,vnatfreq,vt1,vw1,vxi1,vsignal,vdt);
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Deleting all the variables
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
pause(1)
disp(' ')
disp('Deleting all variables from the workspace...')
clear;
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
End of the program
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
pause(1)
disp(' ')
disp('End of Newmark program... Check the results files saved in the specified folder')
207
function STIMAT=f15_tang_stiff_mat_elastoplast_hyster(NDOF,KEY,KELAST)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Newmark method for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the nonlinear tangential stiffness matrix of the
%
current time step for the linear elastoplastic model
%
%
inputs
%
%
NDOF
Number of degrees of freedom of the system
%
KEY
Vector specifying the nonlinear state of each degree of
%
freedom in the previous time step
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=1 Plastic behavior in compression
%
KELAST
Vector containing the elastic stiffness coefficients of
%
the springs
%
%
%
outputs
%
%
STIMAT
Tangential stiffness matrix of the current time step
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
F=zeros(NDOF+1,NDOF);
STIMAT=zeros(NDOF);
if KEY(1)==0
F(1,1)=KELAST(1);
end
if KEY(NDOF+1)==0
F(NDOF+1,NDOF)=KELAST(NDOF+1);
end
if NDOF>1
for IDOF=2:NDOF
if KEY(IDOF)==0
F(IDOF,IDOF1)=KELAST(IDOF);
F(IDOF,IDOF)=KELAST(IDOF);
end
end
end
208
for IDOF=1:NDOF
STIMAT(IDOF,:)=F(IDOF,:)F(IDOF+1,:);
end
209
function STIMAT=f16_tang_stiff_mat_cubic_hyster_v2(NDOF,KEY,KELAST,DISPLA,UO)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Newmark method for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the nonlinear tangential stiffness matrix of the
%
current time step for the cubic elastoplastic model
%
%
inputs
%
%
NDOF
Number of degrees of freedom of the system.
%
KEY
Vector specifying the nonlinear state of each degree of
%
freedom in the previous time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=1 Plastic behavior in compression
%
DISPLA
Displacement vector of the previous time step
%
UO
Zero stress displacement of the previous time step
%
KELAST
Vector containing the elastic stiffness coefficients of
%
the springs
%
%
%
outputs
%
%
STIMAT
Tangential stiffness matrix of the current time step.
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
%Relative displacement
DISREL=zeros(NDOF+1,1);
DISREL(1)=DISPLA(1);
DISREL(NDOF+1)=DISPLA(NDOF);
if NDOF>=2
for IDOF=2:NDOF
DISREL(IDOF)=DISPLA(IDOF)DISPLA(IDOF1);
end
end
% Stiffness matrix coefficients
KTANG=8(32/9)*((DISRELUO).*(DISRELUO));
F=zeros(NDOF+1,NDOF);
STIMAT=zeros(NDOF);
if KELAST(1)==0
F(1,1)=0;
elseif KEY(1)==0  KEY(1)==3
F(1,1)=KTANG(1);
elseif KEY(1)==2
210
F(1,1)=KELAST(1);
end
if KELAST(NDOF+1)==0
F(NDOF+1,NDOF)=0;
elseif KEY(NDOF+1)==0  KEY(NDOF+1)==3
F(NDOF+1,NDOF)=KTANG(NDOF+1);
elseif KEY(NDOF+1)==2
F(NDOF+1,NDOF)=KELAST(NDOF+1);
end
if NDOF>1
for IDOF=2:NDOF
if KELAST(IDOF)==0
F(IDOF,IDOF1)=0;
F(IDOF,IDOF)=0;
elseif KEY(IDOF)==0  KEY(IDOF)==3
F(IDOF,IDOF1)=KTANG(IDOF);
F(IDOF,IDOF)=KTANG(IDOF);
elseif KEY(IDOF)==2
F(IDOF,IDOF1)=KELAST(IDOF);
F(IDOF,IDOF)=KELAST(IDOF);
end
end
end
for IDOF=1:NDOF
STIMAT(IDOF,:)=F(IDOF,:)F(IDOF+1,:);
end
211
function DAMMAT=F17_tang_damp_mat_nonlimet(NDOF,KEY,CELAST,DISPLA)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Newmark method for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the nonlinear tangential damping matrix of the
%
current time step for the quadractic displacement dependent viscous
%
model
%
%
inputs
%
%
NDOF
Number of degrees of freedom of the system.
%
KEY
Vector specifying the nonlinear state of each degree of
%
freedom in the previous time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=1 Plastic behavior in compression
%
DISPLA
Displacement vector of the previous time step
%
UO
Zero stress displacement of the previous time step
%
CELAST
Vector containing the elastic damping coefficients of
%
the dashpots
%
%
%
outputs
%
%
DAMMAT
Tangential damping matrix of the current time step.
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
%Nonlinear damping parameters
alfa=1;
a=zeros(NDOF+1,1);
for IDOF=1:NDOF+1
if CELAST(IDOF)>0
a(IDOF)=alfa;
elseif CELAST(IDOF)<0
a(IDOF)=alfa;
end
end
%Relative displacement
DISREL=zeros(NDOF+1,1);
DISREL(1)=DISPLA(1);
DISREL(NDOF+1)=DISPLA(NDOF);
if NDOF>=2
for IDOF=2:NDOF
DISREL(IDOF)=DISPLA(IDOF)DISPLA(IDOF1);
212
end
end
% Damping matrix coefficients
F=zeros(NDOF+1,NDOF);
DAMMAT=zeros(NDOF);
CTANG=transp(CELAST).*(1+a.*(DISREL.^2));
if KEY(1)==0
F(1,1)=CTANG(1);
end
if KEY(NDOF+1)==0
F(NDOF+1,NDOF)=CTANG(NDOF+1);
end
if NDOF>1
for IDOF=2:NDOF
if KEY(IDOF)==0
F(IDOF,IDOF1)=CTANG(IDOF);
F(IDOF,IDOF)=CTANG(IDOF);
end
end
end
for IDOF=1:NDOF
DAMMAT(IDOF,:)=F(IDOF,:)F(IDOF+1,:);
end
213
function
[DISPLAOUT,VELOCOUT,ACCELEOUT,FNLSTIOUT,FNLDAMOUT,NITERA,KEYSTIOUT,SPRFOROUT,DASHFOUT,UTOUT,UCOUT
,UOOUT,EMKEY]=f18_Newt_Raph_modf_v6(NDOF,DISPLAIN,VELOCIN,ACCELEIN,FNLSTIN,FNLDAMIN,EFVALO,EFFSTI
,M,tol,KEYSTIN,SPRFORIN,DASHFOIN,UTIN,UCIN,UOIN,ITIME,maxite,gama,beta,DT,LODFAC,P,VARLOD,KELAST,
CELAST)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Newmark method for simplified MDOF systems
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Modified Newton Raphson iterative procedure to calculate the nonlinear
%
response of the system. The file containing the nonlinear models for
%
stiffness and damping should be selected below
%
%
inputs
%
%
%
NDOF
Number of degrees of freedom of the system.
%
DISPLAIN
Displacement vector of the previous time step
%
VELOCIN
Velocity vector of the previous time step
%
ACCELEIN
Acceleration vector of the previous time step
%
FNLSTIN
Nonlinear restoring stiffness force vector of the
%
previous time step
%
FNLDAMIN
Nonlinear restoring damping force vector of the
%
previous time step
%
EFVALO
Effective load increment of the current time step
%
EFFSTI
Effective stiffness matrix of the current time step
%
M
Mass matrix of the system
%
tol
Tolerance for convergence criteria
%
KEYSTIN
Vector specifying the stiffness nonlinear state of each
%
degree of freedom in the previous time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=1 Plastic behavior in compression
%
SPRFORIN
Spring internal stiffness force of the previous time
%
step
%
DASHFOIN
Dashpot internal damping force of the previous time
%
step
%
UTIN
Maximum tensile displacement in the elastic range of
%
the previous time step
%
UCIN
Maximum compression displacement in the elastic range
%
of the previous time step
%
UOIN
Zero stress displacement of the previous time step
%
ITIME
Current time step index
%
maxite
Maximum number of iterations
%
gamma
gamma parameter of the Newmark method
%
beta
beta parameter of the Newmark method
%
DT
Time step size
214
%
LODFAC
Load factor of the current time step
%
P
Load vector
%
VARLOD
Load increment of the current time step
%
KELAST
Vector containing the elastic stiffness coefficients of
%
the springs
%
CELAST
Vector containing the elastic damping coefficients of
%
the dashpots
%
%
%
outputs
%
%
DISPLAOUT
Displacement vector of the current time step
%
VELOCOUT
Velocity vector of the current time step
%
ACCELEOUT
Acceleration vector of the current time step
%
FNLSTIOUT
Nonlinear restoring stiffness force vector of the
%
current time step
%
FNLDAMOUT
Nonlinear restoring damping force vector of the
%
current time step
%
NITERA
Number of iterations required to achieve convergence
%
KEYSTIOUT
Vector specifying the stiffness nonlinear state of each
%
degree of freedom in the current time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=1 Plastic behavior in compression
%
SPRFOROUT
Spring internal stiffness force of the current time
%
step
%
DASHFOUT
Dashpot internal damping force of the current time
%
step
%
UTOUT
Maximum tensile displacement in the elastic range of
%
the current time step
%
UCOUT
Maximum compression displacement in the elastic range
%
of the current time step
%
UOOUT
Zero stress displacement of the current time step
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Initialization of variables
u=zeros(NDOF,maxite+1);
fs=zeros(NDOF,maxite+1);
fd=zeros(NDOF,maxite+1);
dr=zeros(NDOF,maxite+1);
du=zeros(NDOF,maxite+1);
u(:,1)=DISPLAIN;
fs(:,1)=FNLSTIN;
fd(:,1)=FNLDAMIN;
dr(:,1)=EFVALO;
kt=EFFSTI;
error=1;
j=0;
215
216
ACCELEOUT=M\(LODFAC*PFNLDAMOUTFNLSTIOUT);
NITERA=j;
217
function
[KEYOUT,INTFOROUT,FNLSTIOUT,UTOUT,UCOUT,UOOUT,EMKEY]=f9_lin_stiff_int_force(NDOF,KEYIN,DISPLA1,DI
SPLA2,INTFORIN,KELAST)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Stiffness force models
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the internal stiffness force corresponding to
%
the linear elastic model
%
%
inputs
%
%
NDOF
Number of degrees of freedom of the system.
%
KEYIN
Vector specifying the nonlinear state of each degree of
%
freedom in the previous time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=1 Plastic behavior in compression
%
DISPLA1
Displacement vector of the previous time step
%
DISPLA2
Displacement vector of the current time step
%
INTFORIN
Spring internal stiffness force of the previous time
%
step
%
KELAST
Vector containing the elastic stiffness coefficients of
%
the springs
%
%
%
outputs
%
%
KEYOUT
Vector speciying the nonlinear state of each degree of
%
freedom in the current time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=1 Plastic behavior in compression
%
INTFOROUT
Spring internal stiffness force of the current time
%
step
%
FNLSIOUT
Nonlinear restoring stiffness force vector of the
%
current time step
%
%
218
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Calculation of relative displacements increments
DISREL1=zeros(NDOF+1,1);
DISREL2=zeros(NDOF+1,1);
DISREL1(1)=DISPLA1(1);
DISREL1(NDOF+1)=DISPLA1(NDOF);
DISREL2(1)=DISPLA2(1);
DISREL2(NDOF+1)=DISPLA2(NDOF);
if NDOF>=2
for IDOF=2:NDOF
DISREL1(IDOF)=DISPLA1(IDOF)DISPLA1(IDOF1);
DISREL2(IDOF)=DISPLA2(IDOF)DISPLA2(IDOF1);
end
end
VARDISREL=DISREL2DISREL1;
%
% Defining the nonlinear state and nonlinear internal spring force for
% each degree of freedom in the current time step
EMKEY=100;
KEYOUT=zeros(NDOF+1,1);
INTFOROUT=zeros(NDOF+1,1);
UTOUT=zeros(NDOF+1,1);
UCOUT=zeros(NDOF+1,1);
UOOUT=zeros(NDOF+1,1);
for IDOF=1:NDOF+1
if KELAST(IDOF)==0
INTFOROUT(IDOF)=0;
elseif KEYIN(IDOF)==0
INTFOROUT(IDOF)=INTFORIN(IDOF)+KELAST(IDOF)*VARDISREL(IDOF);
end
end
%
% Recalculation of the nonlinear restoring force vector for the current
% time step
FNLSTIOUT=zeros(NDOF,1);
for IDOF=1:NDOF
FNLSTIOUT(IDOF)=INTFOROUT(IDOF)INTFOROUT(IDOF+1);
end
219
function
[KEYOUT,INTFOROUT,FNLSTIOUT,UTOUT,UCOUT,UOOUT,EMKEY]=f10_nonlin_hyster_stiff_int_force_v3(VELOCI,
NDOF,KEYIN,DISPLA1,DISPLA2,INTFORIN,UTIN,UCIN,UOIN,ITIME,KELAST)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Stiffness force models
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the internal stiffness force corresponding to
%
the elastoplastic hysteretic model
%
%
inputs
%
%
%
VELOCI
Velocity vector of the current time step.
%
NDOF
Number of degrees of freedom of the system.
%
KEYIN
Vector specifying the nonlinear state of each degree of
%
freedom in the previous time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=1 Plastic behavior in compression
%
DISPLA1
Displacement vector of the previous time step
%
DISPLA2
Displacement vector of the current time step
%
INTFORIN
Spring internal stiffness force of the previous time
%
step
%
UTIN
Maximum tensile displacement in the elastic range of
%
the previous time step
%
UCIN
Maximum compression displacement in the elastic range
%
of the previous time step
%
UOIN
Zero stress displacement of the previous time step
%
ITIME
Current time step index
%
KELAST
Vector containing the elastic stiffness coefficients of
%
the springs
%
%
%
outputs
%
%
KEYOUT
Vector speciying the nonlinear state of each degree of
%
freedom in the current time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=1 Plastic behavior in compression
%
INTFOROUT
Spring internal stiffness force of the current time
%
step
%
FNLSIOUT
Nonlinear restoring stiffness force vector of the
%
current time step
%
UTOUT
Maximum tensile displacement in the elastic range of
%
the current time step
220
%
UCOUT
Maximum compression displacement in the elastic range
%
of the current time step
%
UOOUT
Zero stress displacement of the current time step
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Define the nonlinear elastoplastic parameters for each degree of
% freedom
% This data has to be modified by the user depending of the problem data
%
TENSTR=[7.5 7.5 12.35 12.35 0 0 0 0 0 0];% Tensile strength for each degree of freedom
COMSTR=[7.5 7.5 12.35 12.35 0 0 0 0 0 0];% Compressive strength for each degree of freedom
if ITIME==1
for IDOF=1:NDOF+1
UTIN(IDOF)=TENSTR(IDOF)/KELAST(IDOF);
UCIN(IDOF)=COMSTR(IDOF)/KELAST(IDOF);
end
sizet=size(TENSTR);
sizec=size(COMSTR);
sizeut=size(UTIN);
sizeuc=size(UCIN);
sizeuo=size(UOIN);
%
% Checking if the number of parameters provided are enough
%
if sizet(2)<NDOF+1  sizec(2)<NDOF+1
disp('Error message: the number coefficients of the vectors "TENSTR" and COMSTR, defined
in function "nonlin_hyster_stiff_int_force...", should be equal or higher to NDOF +1')
elseif sizeut(2)<NDOF+1  sizeuc(2)<NDOF+1  sizeuo(2)<NDOF+1
disp('Error message: the number coefficients of the vectors "UTIN", "UCIN" AND "UOIN",
defined in function "nonlin_hyster_stiff_int_force", should be equal or higher to NDOF +1')
end
end
%
% Calculation of relative displacements increments and relative velocities
%
DISREL1=zeros(NDOF+1,1);
DISREL2=zeros(NDOF+1,1);
VELREL=zeros(NDOF+1,1);
DISREL1(1)=DISPLA1(1);
DISREL1(NDOF+1)=DISPLA1(NDOF);
DISREL2(1)=DISPLA2(1);
DISREL2(NDOF+1)=DISPLA2(NDOF);
VELREL(1)=VELOCI(1);
VELREL(NDOF+1)=VELOCI(NDOF);
if NDOF>=2
for IDOF=2:NDOF
DISREL1(IDOF)=DISPLA1(IDOF)DISPLA1(IDOF1);
DISREL2(IDOF)=DISPLA2(IDOF)DISPLA2(IDOF1);
221
VELREL(IDOF)=VELOCI(IDOF)VELOCI(IDOF1);
end
end
VARDISREL=DISREL2DISREL1;
%
% Defining the nonlinear state and nonlinear internal force for each degree
% of freedom in the current time step
EMKEY=100;
KEYOUT=zeros(NDOF+1,1);
INTFOROUT=zeros(NDOF+1,1);
UTOUT=UTIN;
UCOUT=UCIN;
UOOUT=UOIN;
for IDOF=1:NDOF+1
if KELAST(IDOF)==0
INTFOROUT(IDOF)=0;
elseif KEYIN(IDOF)==0
INTFOROUT(IDOF)=INTFORIN(IDOF)+KELAST(IDOF)*VARDISREL(IDOF);
if INTFOROUT(IDOF)>=TENSTR(IDOF)
KEYOUT(IDOF)=1;
INTFOROUT(IDOF)=TENSTR(IDOF);
elseif INTFOROUT(IDOF)<=COMSTR(IDOF)
KEYOUT(IDOF)=1;
INTFOROUT(IDOF)=COMSTR(IDOF);
else
KEYOUT(IDOF)=0;
end
elseif KEYIN(IDOF)==1
INTFOROUT(IDOF)=TENSTR(IDOF);
if VELREL(IDOF)>0
KEYOUT(IDOF)=1;
else
KEYOUT(IDOF)=0;
if VARDISREL(IDOF)<0
INTFOROUT(IDOF)=TENSTR(IDOF)+KELAST(IDOF)*VARDISREL(IDOF);
UTOUT(IDOF)=DISREL1(IDOF);
else
UTOUT(IDOF)=DISREL2(IDOF);
end
UCOUT(IDOF)=UTOUT(IDOF)(TENSTR(IDOF)COMSTR(IDOF))/KELAST(IDOF);
UOOUT(IDOF)=UTOUT(IDOF)TENSTR(IDOF)/KELAST(IDOF);
end
else
INTFOROUT(IDOF)=COMSTR(IDOF);
if VELREL(IDOF)<0
KEYOUT(IDOF)=1;
else
KEYOUT(IDOF)=0;
if VARDISREL(IDOF)>0
INTFOROUT(IDOF)=COMSTR(IDOF)+KELAST(IDOF)*VARDISREL(IDOF);
222
UCOUT(IDOF)=DISREL1(IDOF);
else
UCOUT(IDOF)=DISREL2(IDOF);
end
UTOUT(IDOF)=UCOUT(IDOF)+(TENSTR(IDOF)COMSTR(IDOF))/KELAST(IDOF);
UOOUT(IDOF)=UTOUT(IDOF)TENSTR(IDOF)/KELAST(IDOF);
end
end
end
%
% Recalculation of the nonlinear restoring force vector for the current time
% step
FNLSTIOUT=zeros(NDOF,1);
for IDOF=1:NDOF
FNLSTIOUT(IDOF)=INTFOROUT(IDOF)INTFOROUT(IDOF+1);
end
223
function
[KEYOUT,INTFOROUT,FNLSTIOUT,UTOUT,UCOUT,UOOUT,EMKEY]=f11_nonlin_hyster_stiff_int_force_cubic(VELO
CI,NDOF,KEYIN,DISPLA1,DISPLA2,UTIN,UCIN,UOIN,ITIME,KELAST)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Stiffness force models
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the internal stiffness force corresponding to
%
the cubic elastoplastic hysteretic model
%
%
%
inputs
%
%
VELOCI
Velocity vector of the current time step.
%
NDOF
Number of degrees of freedom of the system.
%
KEYIN
Vector specifying the nonlinear state of each degree of
%
freedom in the previous time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=1 Plastic behavior in compression
%
DISPLA1
Displacement vector of the previous time step
%
DISPLA2
Displacement vector of the current time step
%
UTIN
Maximum tensile displacement in the elastic range of
%
the previous time step
%
UCIN
Maximum compression displacement in the elastic range
%
of the previous time step
%
UOIN
Zero stress displacement of the previous time step
%
ITIME
Current time step index
%
KELAST
Vector containing the elastic stiffness coefficients of
%
the springs
%
%
%
outputs
%
%
KEYOUT
Vector speciying the nonlinear state of each degree of
%
freedom in the current time step.
%
KEY=0
Elastic behavior
%
KEY=1
Plastic behavior in tension
%
KEY=1 Plastic behavior in compression
%
INTFOROUT
Spring internal stiffness force of the current time
%
step
%
FNLSIOUT
Nonlinear restoring stiffness force vector of the
%
current time step
%
UTOUT
Maximum tensile displacement in the elastic range of
%
the current time step
%
UCOUT
Maximum compression displacement in the elastic range
%
of the current time step
224
%
UOOUT
Zero stress displacement of the current time step
%
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Define the nonlinear cubic elastoplastic parameters for each degree of
% freedom
% Cubic nonlinearity
fs=k1*uk3*u^3
% This data has to be modified by the user depending of the problem data
%
k1=[8 8 8 8 8 8 8 8 8 8];
k3=[32/27 32/27 32/27 32/27 32/27 32/27 32/27 32/27 32/27 32/27];
d=sqrt(k1./(3*k3));
TENSTR=(k1.*d)k3.*(d.^3);% Tensile strength for each degree of freedom
COMSTR=TENSTR;% Compressive strength for each degree of freedom
if ITIME==1
UTIN=transp(d);
UCIN=UTIN;
sizet=size(TENSTR);
sizec=size(COMSTR);
sizeut=size(UTIN);
sizeuc=size(UCIN);
sizeuo=size(UOIN);
%
% Checking if the number of parameters provided are enough
%
if sizet(2)<NDOF+1  sizec(2)<NDOF+1
disp('Error message: the number coefficients of the vectors "TENSTR" and COMSTR, defined
in function "nonlin_hyster_stiff_int_force...", should be equal or higher to NDOF +1')
elseif sizeut(2)<NDOF+1  sizeuc(2)<NDOF+1  sizeuo(2)<NDOF+1
disp('Error message: the number coefficients of the vectors "UTIN", "UCIN" AND "UOIN",
defined in function "nonlin_hyster_stiff_int_force", should be equal or higher to NDOF +1')
end
end
%
% Calculation of relative displacements increments and relative velocities
DISREL1=zeros(NDOF+1,1);
DISREL2=zeros(NDOF+1,1);
VELREL=zeros(NDOF+1,1);
DISREL1(1)=DISPLA1(1);
DISREL1(NDOF+1)=DISPLA1(NDOF);
DISREL2(1)=DISPLA2(1);
DISREL2(NDOF+1)=DISPLA2(NDOF);
VELREL(1)=VELOCI(1);
VELREL(NDOF+1)=VELOCI(NDOF);
if NDOF>=2
for IDOF=2:NDOF
DISREL1(IDOF)=DISPLA1(IDOF)DISPLA1(IDOF1);
DISREL2(IDOF)=DISPLA2(IDOF)DISPLA2(IDOF1);
225
VELREL(IDOF)=VELOCI(IDOF)VELOCI(IDOF1);
end
end
VARDISREL=DISREL2DISREL1;
%
% Defining the nonlinear state and nonlinear internal force for each degree
% of freedom in the current time step
EMKEY=100; %Value of state variable for error message
KEYOUT=zeros(NDOF+1,1);
INTFOROUT=zeros(NDOF+1,1);
UTOUT=UTIN;
UCOUT=UCIN;
UOOUT=UOIN;
for IDOF=1:NDOF+1
if KELAST(IDOF)==0
INTFOROUT(IDOF)=0;
elseif KEYIN(IDOF)==0
if DISREL2(IDOF)>=UTIN(IDOF)
KEYOUT(IDOF)=1;
INTFOROUT(IDOF)=TENSTR(IDOF);
elseif DISREL2(IDOF)<=UCIN(IDOF)
KEYOUT(IDOF)=1;
INTFOROUT(IDOF)=COMSTR(IDOF);
else
KEYOUT(IDOF)=0;
INTFOROUT(IDOF)=k1(IDOF)*(DISREL2(IDOF)UOIN(IDOF))k3(IDOF)*((DISREL2(IDOF)UOIN(IDOF))^3);
end
elseif KEYIN(IDOF)==1
INTFOROUT(IDOF)=TENSTR(IDOF);
if VELREL(IDOF)>0
KEYOUT(IDOF)=1;
else
KEYOUT(IDOF)=2;
if VARDISREL(IDOF)<0
INTFOROUT(IDOF)=TENSTR(IDOF)+KELAST(IDOF)*VARDISREL(IDOF);
UTOUT(IDOF)=DISREL1(IDOF);
else
UTOUT(IDOF)=DISREL2(IDOF);
end
UOOUT(IDOF)=UTOUT(IDOF)TENSTR(IDOF)/KELAST(IDOF);
UCOUT(IDOF)=UOOUT(IDOF)d(IDOF);
end
elseif KEYIN(IDOF)==2
if DISREL2(IDOF)<=UOIN(IDOF)
KEYOUT(IDOF)=3;
INTFOROUT(IDOF)=k1(IDOF)*(DISREL2(IDOF)UOIN(IDOF))k3(IDOF)*((DISREL2(IDOF)UOIN(IDOF))^3);
elseif DISREL2(IDOF)>= UTIN(IDOF)
KEYOUT(IDOF)=1;
226
INTFOROUT(IDOF)=TENSTR(IDOF);
else
KEYOUT(IDOF)=2;
INTFOROUT(IDOF)=KELAST(IDOF)*(DISREL2(IDOF)UOIN(IDOF));
end
elseif KEYIN(IDOF)==3
if DISREL2(IDOF)<=UCIN(IDOF)
KEYOUT(IDOF)=1;
INTFOROUT(IDOF)=COMSTR(IDOF);
elseif DISREL2(IDOF)>= UOIN(IDOF)
KEYOUT(IDOF)=2;
INTFOROUT(IDOF)=KELAST(IDOF)*(DISREL2(IDOF)UOIN(IDOF));
else
KEYOUT(IDOF)=3;
INTFOROUT(IDOF)=k1(IDOF)*(DISREL2(IDOF)UOIN(IDOF))k3(IDOF)*((DISREL2(IDOF)UOIN(IDOF))^3);
end
else
EMKEY=KEYIN(IDOF);
SPRINGNUM=IDOF;
fprintf('\nNot defined state variable KEYIN=%d of spring number %d for the calculation of
time step %d\n',EMKEY,SPRINGNUM,ITIME+1)
return
end
end
%
% Recalculation of the nonlinear restoring force vector for the current time
% step
FNLSTIOUT=zeros(NDOF,1);
for IDOF=1:NDOF
FNLSTIOUT(IDOF)=INTFOROUT(IDOF)INTFOROUT(IDOF+1);
end
227
function [FNLDAMOUT,INTFOROUT]=f12_lin_damp_int_force_v2(NDOF,VARVEL,INTFORIN,CELAST)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Damping force models
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the internal damping force corresponding to
%
the linear model
%
%
inputs
%
%
NDOF
Number of degrees of freedom of the system
%
VARVEL
Velocity increment vector from the previous to the
%
current time step
%
INTFORIN
Dashpot internal damping force of the previous time
%
step
%
CELAST
Vector containing the linear damping coefficients of
%
the dashpots
%
%
%
outputs
%
%
INTFOROUT
Dashpot internal damping force of the current time
%
step
%
FNLDAMOUT
Nonlinear restoring damping force vector of the
%
current time step
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Calculating the relative variation of velocity
VARVELREL=zeros(NDOF+1,1);
VARVELREL(1)=VARVEL(1);
VARVELREL(NDOF+1)=VARVEL(NDOF);
if NDOF>=2
for IDOF=2:NDOF
VARVELREL(IDOF)=VARVEL(IDOF)VARVEL(IDOF1);
end
end
%
% Calculation of dashpot force
INTFOROUT=zeros(NDOF+1,1);
for IDOF=1:NDOF+1
INTFOROUT(IDOF)=INTFORIN(IDOF)+CELAST(IDOF)*VARVELREL(IDOF);
228
end
%
% Calculation of nonlinear internal damping force
FNLDAMOUT=zeros(NDOF,1);
for IDOF=1:NDOF
FNLDAMOUT(IDOF)=INTFOROUT(IDOF)INTFOROUT(IDOF+1);
end
229
function [FNLDAMOUT,INTFOROUT]=f13_nonlinmet_damp_int_force(NDOF,DISPLA,VELOCI,CELAST)
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
Damping force models
%
L. Fernando Sirumbal Z.
%
April, 2013
%
%
Function to calculate the internal damping force corresponding to
%
the nonlinear displacement dependent model
%
%
inputs
%
%
NDOF
Number of degrees of freedom of the system
%
DISPLA
Displacement vector of the current time step
%
VELOCI
Velocity vector of the current time step
%
CELAST
Vector containing the linear damping coefficients of
%
the dashpots
%
%
%
outputs
%
%
INTFOROUT
Dashpot internal damping force of the current time
%
step
%
FNLDAMOUT
Nonlinear restoring damping force vector of the
%
current time step
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Quadratic displacement dependent nonlinearity
%
fd=CELAST*VELOCI(ALFA*DISPLA^2+1)
%
%Nonlinear damping parameters
alfa=1;
%
%Relative displacement
DISREL=zeros(NDOF+1,1);
DISREL(1)=DISPLA(1);
DISREL(NDOF+1)=DISPLA(NDOF);
if NDOF>=2
for IDOF=2:NDOF
DISREL(IDOF)=DISPLA(IDOF)DISPLA(IDOF1);
end
end
%
%Relative velocity
VELREL=zeros(NDOF+1,1);
VELREL(1)=VELOCI(1);
VELREL(NDOF+1)=VELOCI(NDOF);
230
if NDOF>=2
for IDOF=2:NDOF
VELREL(IDOF)=VELOCI(IDOF)VELOCI(IDOF1);
end
end
%
% Calculation of dashpot force
INTFOROUT=zeros(NDOF+1,1);
for IDOF=1:NDOF+1
INTFOROUT(IDOF)=VELREL(IDOF)*CELAST(IDOF)*(alfa*((DISREL(IDOF))^2)+1);
%INTFOROUT(IDOF)=VELREL(IDOF)*CELAST(IDOF)*(alfa*((DISREL(IDOF))^2)1);
end
%
% Calculation of nonlinear internal damping force
FNLDAMOUT=zeros(NDOF,1);
for IDOF=1:NDOF
FNLDAMOUT(IDOF)=INTFOROUT(IDOF)INTFOROUT(IDOF+1);
end
231
232
*FILOS
INITIA
*INPUT
*PHASE
BEGIN ACTIVE
ELEMEN "Dam" "Found" /
REINFO
END ACTIVE
*NONLIN
BEGIN EXECUT
BEGIN LOAD
LOADNR 1
STEPS EXPLIC SIZES 0.25(4)
END LOAD
ITERAT CONVER SIMULT
END EXECUT
*PHASE
*HFTD
BEGIN EIGEN
EXECUT NMODES 15
BEGIN OUTPUT
FXPLUS
FILE "eigenHFTD_dry"
END OUTPUT
END EIGEN
BEGIN MODEL
BEGIN MATRIX
DAMPIN
STRESS PHASE
END MATRIX
END MODEL
BEGIN RESPON
BEGIN EXECUT
ITERAT MAXITE 50
MAXFRE 25
TIMESE STEPS EXPLIC NUMBER 50
END EXECUT
BEGIN OUTPUT
FXPLUS
FILE "respHFTD_wet"
ACCELE TOTAL TRANSL GLOBAL
DISPLA TOTAL TRANSL GLOBAL
FORCE REACTI TRANSL GLOBAL
FSPRES TOTAL
STATUS CRACK
STRAIN TOTAL GREEN GLOBAL
STRAIN CRACK GREEN
STRESS TOTAL CAUCHY GLOBAL
VELOCI TOTAL TRANSL GLOBAL
END OUTPUT
REDUCE SELECT MODES 115 /
END RESPON
*END
233
Lebih dari sekadar dokumen.
Temukan segala yang ditawarkan Scribd, termasuk buku dan buku audio dari penerbitpenerbit terkemuka.
Batalkan kapan saja.